Method and device for determining reasoning result

Abstract

The invention relates to the technical field of natural language processing, in particular to a reasoning result determination method and device. The reasoning result determination method comprises the steps that a target problem is acquired, and the target problem comprises a premise set and a target hypothesis; according to the score corresponding to any determined premise in a first premise subset, a second premise subset is obtained, and the first premise subset comprises at least one determined premise in the premise set; obtaining the inference premise which passes verification and corresponds to the second premise subset, and adding the inference premise to the premise set; under the condition that the iteration adding operation of the premise set meets a preset iteration requirement, performing question operation on the iterated premise set and the target hypothesis, and obtaining a question operation result; and according to the questioning operation result, determining a reasoning result of the target question. By adopting the method and the device, the reasoning efficiency and the reasoning accuracy can be improved under the condition that the reasoning cost is not increased.

Description

Technical Field

[0001] The present disclosure relates to the technical field of natural language processing, and in particular to a method and device for determining an inference result. Background Art

[0002] With the development of science and technology, natural language processing has been applied to various fields. For example, it can be applied to the field of prompt engineering to perform reasoning on logical tasks. For example, cumulative reasoning can be used to perform reasoning on logical tasks, or external formal logic tools can be used to perform reasoning on logical tasks. However, the reasoning cost will be increased in this process, and cumulative reasoning will result in blind reasoning, which will lead to low reasoning efficiency and accuracy. Summary of the invention

[0003] The present disclosure provides a method and device for determining an inference result, so as to improve the inference efficiency and inference accuracy without increasing the inference cost. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided a method for determining an inference result, including:

[0005] Acquire a target problem, wherein the target problem includes a premise set and a target hypothesis, the premise set includes at least one premise, the premise includes condition information corresponding to the target problem, and the target hypothesis is a hypothesis to be determined proposed for the premise set;

[0006] According to the score corresponding to any determined premise in the first premise subset, a second premise subset is obtained, wherein the first premise subset includes at least one determined premise in the premise set, and the second premise subset includes at least one determined premise and / or at least one undetermined premise in the premise set, the determined premise is condition information related to the target hypothesis and does not carry a selection format, and the undetermined premise is condition information that is unrelated to the target hypothesis or carries a selection format;

[0007] Acquire a verified inference premise corresponding to the second premise subset, and add the inference premise to the premise set, wherein the inference premise is a premise obtained by inferring the second premise subset;

[0008] When the iterative addition operation of the premise set meets the preset iteration requirement, performing a questioning operation on the premise set after the iteration and the target hypothesis to obtain a questioning operation result;

[0009] According to the questioning operation result, the inference result of the target question is determined.

[0010] According to some embodiments, obtaining the second premise subset according to a score corresponding to any determined premise in the first premise subset includes:

[0011] Determining a target determination premise according to a first score of any premise in the first premise subset;

[0012] selecting at least one premise that meets the score requirement according to a second score of any premise in the premise subset except the target-determined premise;

[0013] The target determination prerequisite and the at least one prerequisite that satisfies the score requirement are added to a second subset of prerequisites.

[0014] According to some embodiments, determining the target determination premise according to the first score of any premise in the first premise subset includes:

[0015] Determine a first score corresponding to any determined premise in the first premise subset according to the degree of association and the logical relationship between the target hypothesis and the determined premise;

[0016] According to the first score corresponding to any of the determined premises, taking the determined premise with the highest first score in the first premise subset as the target determined premise;

[0017] Before selecting at least one premise that meets the score requirement according to the second score of any premise in the premise subset except the target-determined premise, the method further includes:

[0018] According to the degree of association and the logical relationship between any premise in the premise set except the target determination premise and the target determination premise, a second score corresponding to the any premise is determined.

[0019] According to some embodiments, before obtaining the second premise subset according to the score corresponding to any determined premise in the first premise subset, the method further includes:

[0020] The premise set is classified by using a large language model to obtain a first premise subset and a second premise subset, wherein the first premise subset includes at least one definite premise, and the third premise subset includes at least one indefinite premise.

[0021] According to some embodiments, obtaining the verified reasoning premises corresponding to the second premise subset includes:

[0022] Merging and reasoning the second premise subset to obtain reasoning premises corresponding to the second premise subset;

[0023] At least one verification operation is performed on the reasoning premise, and when the verification results corresponding to the at least one verification operation are all verified to be passed, it is determined that the verified reasoning premise is obtained.

[0024] According to some embodiments, the method further comprises:

[0025] In the case that any verification result corresponding to the at least one verification operation fails the verification, the verification failure reason corresponding to the second premise subset and the reasoning premise is recorded.

[0026] According to some embodiments, performing at least one verification operation on the reasoning premise includes at least one of the following:

[0027] Logically verify the premise of the reasoning;

[0028] Conduct source checking and verification of the stated reasoning premises;

[0029] Repeatable verification of the reasoning premise;

[0030] Verifying the validity of the reasoning premise;

[0031] The boundary conditions of the reasoning premise are verified.

[0032] According to some embodiments, when the iterative addition operation of the premise set meets a preset iteration requirement, performing a questioning operation on the premise set after the iteration and the target hypothesis to obtain a questioning operation result includes:

[0033] When the number of verification failures corresponding to the premise set reaches a number threshold, performing a questioning operation on the iterated premise set and the target hypothesis to obtain a questioning operation result; or

[0034] When the number of inference premises corresponding to the premise set reaches a quantity threshold, a question operation is performed on the iterated premise set and the target hypothesis to obtain a question operation result.

[0035] According to some embodiments, the method further comprises:

[0036] When the reasoning result of the target question is not determined according to the questioning operation result, supplementary reasoning is performed on the target question according to the reasoning process corresponding to the target question to obtain the supplementary reasoning result corresponding to the target question, wherein the reasoning process includes at least one of an iterative process, a verification process and a questioning process.

[0037] According to a second aspect of an embodiment of the present disclosure, there is provided an inference result determination device, including:

[0038] A question acquisition unit, which acquires a target question, wherein the target question includes a premise set and a target hypothesis;

[0039] A set acquisition unit, configured to acquire a second premise subset according to a score corresponding to any determined premise in the first premise subset, wherein the first premise subset includes at least one determined premise in the premise set;

[0040] an inference premise adding unit, configured to obtain a verified inference premise corresponding to the second premise subset, and add the inference premise to the premise set;

[0041] A result acquisition unit is used to perform a questioning operation on the premise set after the iteration and the target hypothesis to obtain a questioning operation result when the iterative addition operation of the premise set meets the preset iteration requirement;

[0042] The result determination unit is used to determine the inference result of the target question according to the question operation result.

[0043] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0044] processor;

[0045] a memory for storing instructions executable by the processor;

[0046] The processor is configured to execute the instructions to implement the method for determining the inference result as described in any one of the aforementioned aspects.

[0047] According to a fourth aspect of the present application, a storage medium is provided. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method for determining an inference result described in any one of the aforementioned aspects.

[0048] According to a fifth aspect of the present application, a computer program product is provided, comprising a computer program, wherein the computer program implements any one of the methods described in the preceding aspects when executed by a processor.

[0049] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0050] In some or related embodiments, by obtaining a target problem, wherein the target problem includes a premise set and a target hypothesis; according to the score corresponding to any determined premise in the first premise subset, obtaining a second premise subset, wherein the first premise subset includes at least one determined premise in the premise set; obtaining the verified reasoning premise corresponding to the second premise subset, and adding the reasoning premise to the premise set; when the iterative addition operation of the premise set meets the preset iteration requirements, performing a question operation on the premise set and the target hypothesis after the iteration to obtain the question operation result; according to the question operation result, determining the reasoning result of the target problem. Therefore, without using external tools, the reasoning structure corresponding to the target problem can be obtained, and the reasoning structure can be adjusted by reasoning according to the determined premise, so as to improve the matching of the reasoning result and the reasoning process, reduce the mismatch between the reasoning result and the reasoning process, and make the reasoning process complicated, and the premise can be quantified by scores to improve the iteration efficiency, so as to improve the reasoning efficiency and reasoning accuracy without increasing the reasoning cost.

[0051] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0053] Figure 1 A background schematic diagram showing a method for determining an inference result according to an embodiment of the present disclosure;

[0054] Figure 2 is a flow chart of a method for determining an inference result according to an exemplary embodiment;

[0055] Figure 3 is an example of a reasoning model shown according to an exemplary embodiment;

[0056] Figure 4 is a block diagram of a device for determining an inference result according to an exemplary embodiment;

[0057] Figure 5 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0058] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0059] The embodiments of the present disclosure propose a method and device for determining an inference result. In some embodiments, the terms such as the method for determining an inference result can be replaced with the information processing method, the communication method, etc., the terms such as the device for determining an inference result can be replaced with the information processing device, the communication device, etc., and the terms such as the information processing system, the communication system, etc. can be replaced with each other.

[0060] The embodiments of the present disclosure are not exhaustive, but are only illustrative of some embodiments, and are not intended to be a specific limitation on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementation methods in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined, for example, some or all of the steps of different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0061] In each embodiment of the present disclosure, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between the embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form a new embodiment based on their internal logical relationships.

[0062] The terms used in the embodiments of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0063] In the embodiments of the present disclosure, unless otherwise specified, elements expressed in the singular form, such as "a", "an", "the", "above", "said", "aforementioned", "this", etc., may mean "one and only one", or "one or more", "at least one", etc. For example, when using articles such as "a", "an", "the" in English in translation, the noun after the article may be understood as a singular expression or a plural expression.

[0064] In the embodiments of the present disclosure, “plurality” refers to two or more.

[0065] In some embodiments, the terms “at least one,” “one or more,” “a plurality of,” “multiple,” etc. may be used interchangeably.

[0066] In some embodiments, "at least one of A and B", "A and / or B", "A in one case, B in another case", "in response to one case A, in response to another case B", etc., may include the following technical solutions according to the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). When there are more branches such as A, B, C, etc., the above is also similar.

[0067] In some embodiments, the recording method of "A or B" may include the following technical solutions according to the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). When there are more branches such as A, B, C, etc., the above is also similar.

[0068] The prefixes such as "first" and "second" in the embodiments of the present disclosure are only used to distinguish different description objects, and do not constitute restrictions on the position, order, priority, quantity or content of the description objects. The statement of the description object refers to the description in the context of the claims or embodiments, and should not constitute unnecessary restrictions due to the use of prefixes. For example, if the description object is a "field", the ordinal number before the "field" in the "first field" and the "second field" does not limit the position or order between the "fields", and the "first" and "second" do not limit whether the "fields" they modify are in the same message, nor do they limit the order of the "first field" and the "second field". For another example, if the description object is a "level", the ordinal number before the "level" in the "first level" and the "second level" does not limit the priority between the "levels". For another example, the number of description objects is not limited by the ordinal number, and can be one or more. Taking the "first device" as an example, the number of "devices" can be one or more. In addition, the objects modified by different prefixes may be the same or different. For example, if the description object is "device", then the "first device" and the "second device" may be the same device or different devices, and their types may be the same or different. For another example, if the description object is "information", then the "first information" and the "second information" may be the same information or different information, and their contents may be the same or different.

[0069] In some embodiments, “including A”, “comprising A”, “used to indicate A”, and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0070] In some embodiments, terms such as "in response to ...", "in response to determining ...", "in the case of ...", "at the time of ...", "when ...", "if ...", "if ...", etc. can be used interchangeably.

[0071] In some embodiments, terms such as "greater than", "greater than or equal to", "not less than", "more than", "more than or equal to", "not less than", "higher than", "higher than or equal to", "not lower than", and "above" can be replaced with each other, and terms such as "less than", "less than or equal to", "not greater than", "less than", "less than or equal to", "no more than", "lower than", "lower than or equal to", "not higher than", and "below" can be replaced with each other.

[0072] In some embodiments, devices and equipment may be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they may also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "subject", etc.

[0073] In some embodiments, "network" can be interpreted as devices included in the network, such as access network equipment, core network equipment, etc.

[0074] In some embodiments, "access network device (AN device)" may also be referred to as "radio access network device (RAN device)", "base station (BS)", "radio base station (radio base station)", "fixed station (fixed station)", and in some embodiments may also be understood as "node (node)", "access point (access point)", "transmission point (TP)", "reception point (reception point, RP)", "transmission and / or reception point (transmission / reception point, TRP)"", "panel (panel)", "antenna panel (antenna panel)", "antenna array (antenna array)"", "cell (cell)", "macro cell (macro cell)", "small cell (small cell)", "femto cell (femto cell)", "pico cell (pico cell)", "sector (sector)", "cell group (cell group)", "serving cell (serving cell)", "carrier (carrier)", "component carrier (component carrier)", "bandwidth part (bandwidth part, BWP)" and the like.

[0075] In some embodiments, "terminal" or "terminal device" can be referred to as "user equipment (UE)", "user terminal" "mobile station (MS)", "mobile terminal (MT)", subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, etc.

[0076] In some embodiments, data, information, etc. may be obtained with the user's consent.

[0077] In addition, each element, each row, or each column in the table of the embodiments of the present disclosure may be implemented as an independent embodiment, and the combination of any elements, any rows, and any columns may also be implemented as an independent embodiment.

[0078] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0079] According to some embodiments, large language models used to solve logical reasoning tasks can be roughly divided into three categories:

[0080] 1) Based on fine-tuning methods, the reasoning ability of large language models (LLMs) is optimized by fine-tuning or training specialized modules;

[0081] 2) Based on the contextual learning method, special prompts can be designed to activate the step-by-step reasoning ability of LLM. Both of the above methods directly reason through natural language (NL), which provides greater flexibility than symbolic reasoning. However, the inherent complexity and ambiguity of natural language also brings problems such as unreliable reasoning and hallucinations.

[0082] 3) Methods based on external formal logic tools can use symbolic language as the basic unit of reasoning, which effectively transfers the burden of complex and precise reasoning from LLM to more reliable and interpretable external symbolic solvers. However, symbolic representation is difficult to understand for large language models (LLMs); the use of external tools makes it difficult to ensure the flexibility of the model; since the conditions of such data sets cannot be directly converted into conjunction and disjunction relations, in-depth information mining is required. In addition, such methods can only solve pure logic problems. Comprehensive logical reasoning, such as the analytical reasoning problems involved in LogiQA, is beyond their scope of solution.

[0083] For example, cumulative reasoning CR (Cumulative Reasoning), logical large language model Logic-LLM (Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning), and thought tree ToT (Tree-Of-Thought) can be used for reasoning.

[0084] Cumulative Reasoning (CR) uses the accumulation and iteration of language models to simulate the human thinking process. By breaking down tasks into smaller components, CR simplifies the problem-solving process, making it easier to manage and verify. It draws on two ideas of ToT: first, after completing the reasoning, the reasoning premise is verified, and the original conditions and verified premises are used as the conditions for the final question; second, when searching the solution space, verification takes precedence over expansion to avoid invalid node visits.

[0085] The Tree-Of-Thought (ToT) predefines the solution space of the problem as a tree structure. On this basis, some tree search algorithms such as DFS (depth-first search), BFS (breadth-first search), MCTS (Monte Carlo Tree Search), etc. can be used to find the optimal path to obtain the answer in the solution space, combined with the status evaluation and pruning operations of the tree nodes.

[0086] Logic-LLM (Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning) integrates LLM symbolic reasoning to improve the ability to solve logical problems. This technology first uses LLM to convert natural language problems into a symbolic formula, and then uses a deterministic symbolic solver to perform reasoning on the formulated problem.

[0087] As the number of inference layers increases, the number of nodes that need to be verified increases exponentially. Only by continuously accumulating information can we finally get the answer. The blind search strategy is not only a waste of resources, but more importantly, it may not be able to get the answer to some complex problems (ProofWriter), which makes the inference efficiency and accuracy low.

[0088] Among them, the core limitation of Logic-LLM is the use of external tools, which makes LLM only participate in the previous process. Symbolic logic is too difficult for LLM to understand; in addition, formal logic tools limit the problems that the model can solve to the field of logic. Among them, the core problem of ToT is that it presupposes the solution space structure required to solve the problem. In addition to the possibility of not being able to obtain the correct answer, the presupposed structure will also waste a lot of computing resources for ineffective exploration, thus making the accuracy and efficiency of reasoning low.

[0089] Figure 1 is a flow chart of a method for determining an inference result according to an exemplary embodiment. Figure 1 As shown, the reasoning result determination method can be used in a reasoning result determination scenario, and includes the following steps:

[0090] In step S11, a target question is obtained;

[0091] In some embodiments, the technical solution of the embodiment of the present disclosure may be, for example, a large model logic reasoning optimization method (DetermLR) based on conditional determinism. The technical solution of the embodiment of the present disclosure belongs to the prompt engineering field of natural language processing (NLP).

[0092] According to some embodiments, the target problem refers to, for example, a problem for reasoning. The target problem does not specifically refer to a fixed problem. The target problem may be, for example, a problem in a problem set. The target problem includes a premise set and a target hypothesis. For example, when the premise set in the target problem changes, the target problem may also change accordingly. For example, when the target hypothesis in the target problem changes, the target problem may also change accordingly.

[0093] According to some embodiments, the premise set may be, for example, a collection of at least one premise. The premise set does not specifically refer to a fixed set. For example, when any premise included in the premise set changes, the premise set may also change accordingly. For example, when the number of premises corresponding to the premise set changes, the premise set may also change accordingly.

[0094] In one embodiment of the present disclosure, the premise includes condition information corresponding to the target problem. The target hypothesis is a hypothesis to be determined proposed for the premise set.

[0095] According to some embodiments, in the process of acquiring the inference result, for example, a target problem may be acquired, wherein the target problem includes a premise set and a target hypothesis.

[0096] In step S12, a second premise subset is obtained according to a score corresponding to any determined premise in the first premise subset;

[0097] The first premise subset includes at least one determined premise in the premise set. The first premise subset does not specifically refer to a fixed subset. For example, when the number of determined premises included in the first premise subset changes, the first premise subset may also change accordingly. For example, when the determined premise in the premise set changes, the first premise subset may also change accordingly.

[0098] The determined premise may be, for example, a premise related to the target hypothesis. The undetermined premise may be, for example, a premise unrelated to the target hypothesis. The determination of the relevance may be, for example, determined based on the text of the premise and the text of the target hypothesis. The embodiments of the present disclosure are not limited to this.

[0099] According to some embodiments, the score corresponding to the determination premise may be determined, for example, based on the information included in the determination premise. The score corresponding to the determination premise does not specifically refer to a fixed score. For example, when the score determination method changes, the score may also change accordingly. For example, when the determination premise changes, the score corresponding to the determination premise may also change accordingly.

[0100] The second premise subset may be, for example, a set consisting of some or all of the determined premises in the first premise subset. The second premise subset does not specifically refer to a fixed set. For example, when the number of premises corresponding to the second premise subset changes, the second premise subset may also change accordingly. For example, when any determined premise included in the second premise subset changes, the second premise subset may also change accordingly.

[0101] In one embodiment of the present disclosure, the second premise subset includes at least one definite premise and / or at least one indefinite premise in the premise set. The definite premise is condition information related to the target hypothesis and without a selection format, and the indefinite premise is condition information that is not related to the target hypothesis or with a selection format. For example, the indefinite premise may be condition information including the keyword "if".

[0102] The condition information of the selection format includes, but is not limited to, condition information with assumptions and condition information including selection keywords, etc. The condition information related to the target assumption may be directly related to a noun, an adjective, a subject or an object, for example. The embodiments of the present disclosure do not limit this.

[0103] According to some embodiments, when the premise set is obtained, the second premise subset may be obtained according to the score corresponding to at least one determined premise included in the first premise subset in the premise set. For example, the first premise subset in the premise set may be obtained, at least one determined premise in the first premise subset may be obtained, and the score of any determined premise in the at least one determined premise may be obtained. The second premise subset may be obtained according to the score.

[0104] In step S13, the verified inference premise corresponding to the second premise subset is obtained, and the inference premise is added to the premise set;

[0105] According to some embodiments, the verified reasoning premise can be used to indicate, for example, to verify the reasoning premise obtained from the second premise subset and verify the passed reasoning premise. The verified reasoning premise does not specifically refer to a fixed reasoning premise. For example, when the way of obtaining the reasoning premise changes, the verified reasoning premise can also change accordingly. For example, when the verification method changes, the verified reasoning premise can also change accordingly.

[0106] According to some embodiments, the verified inference premises corresponding to the second premise subset are obtained, and the inference premises are added to the premise set.

[0107] In step S14, when the iterative addition operation of the premise set meets the preset iteration requirements, a question operation is performed on the premise set and the target hypothesis after the iteration to obtain the question operation result;

[0108] According to some embodiments, the iterative addition operation of the premise set may be, for example, repeatedly executing step S12 and step S13. The preset iteration requirement does not specifically refer to a fixed requirement. For example, the preset iteration requirement may include the number of iterations or the number of inference premises that failed to be verified. The embodiments of the present disclosure do not limit this.

[0109] In some embodiments, the question operation result may be, for example, the result obtained after the question operation is performed on the premise set and the target hypothesis after iteration. The question operation may be performed on the premise set and the target hypothesis after iteration, for example, at least once. The question operation result, for example, does not specifically refer to a fixed result. For example, when the number of operations corresponding to the question operation changes, the question operation result may also change accordingly. For example, when the premise set after iteration changes, the question operation result may also change accordingly.

[0110] According to some embodiments, the premise set after iteration may include, for example, the original premise set of the target problem and at least one verified inference premise. Among them, the at least one verified inference premise in the premise set after iteration may be used as a premise.

[0111] According to some embodiments, when it is obtained that the iterative addition operation on the premise set meets the preset iteration requirement, a questioning operation is performed on the premise set and the target hypothesis after the iteration to obtain a questioning operation result.

[0112] In step S15, the inference result of the target question is determined according to the question operation result.

[0113] According to some embodiments, after obtaining the question operation result, the inference result of the target question can be determined according to the question operation result.

[0114] In some or related embodiments, by obtaining a target problem, wherein the target problem includes a premise set and a target hypothesis; according to the score corresponding to any determined premise in the first premise subset, obtaining a second premise subset, wherein the first premise subset includes at least one determined premise in the premise set; obtaining the verified reasoning premise corresponding to the second premise subset, and adding the reasoning premise to the premise set; when the iterative addition operation of the premise set meets the preset iteration requirements, performing a question operation on the premise set and the target hypothesis after the iteration to obtain the question operation result; according to the question operation result, determining the reasoning result of the target problem. Therefore, without using external tools, the reasoning structure corresponding to the target problem can be obtained, and the reasoning structure can be adjusted by reasoning according to the determined premise, so as to improve the matching of the reasoning result and the reasoning process, reduce the mismatch between the reasoning result and the reasoning process, and make the reasoning process complicated, and the premise can be quantified by scores to improve the iteration efficiency, so as to improve the reasoning efficiency and the reasoning accuracy without increasing the reasoning cost.

[0115] Figure 2 is a flow chart of a method for determining an inference result according to an exemplary embodiment. Figure 2 As shown, the reasoning result determination method can be used in a reasoning result determination scenario, and includes the following steps:

[0116] In step S21, a target question is obtained;

[0117] Among them, the target problem includes a set of premises and a target hypothesis.

[0118] The specific process is as described above and will not be repeated here.

[0119] According to some embodiments, the technical solution of the present disclosure can be applied to the fields of mathematics, logic, and other fields.

[0120] In step S22, the premise set is classified using a large language model to obtain a first premise subset and a third premise subset;

[0121] The first premise subset includes at least one definite premise, and the third premise subset includes at least one indefinite premise.

[0122] The large prediction model may be, for example, a logic large language model Logic-LLM (Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning). Figure 3The model includes input premises, input target, premise identifier, and premise determination.

[0123] Determinate premises; Indeterminate premises; Premise identification module; Most relevant premise; Candidate premises; Supplement Scorer; Selected premises: Proposition explorer; New proposition; Logic Verifier; Relevance Scorer; Premiseprioritization and exploration; Memory extraction; Memory update; Iterative process with reasoning memory; Memory storage; Loop judgement.

[0124] Among them, the number of determined premises in the premise identification module can be, for example, less than or equal to the number of determined premises after loop judgment, and the number of undetermined premises in the premise identification module can be, for example, greater than or equal to the number of undetermined premises after loop judgment.

[0125] According to some embodiments, the large language model can, for example, be classified according to the text parts corresponding to each premise in the premise subset. For example, classification can be performed according to the first text of each premise and the second text of the target hypothesis. For example, whether the premise is a definite premise can be determined based on whether the premise is directly related to the target hypothesis, whether it is a simple proposition, whether it is a unary assignment expression, etc. For example, classification can be performed based on the correlation between the first text of each premise and the second text of the target hypothesis. Among them, for example, when the same text exists in the first text and the second text in the premise, the premise can be added to the first premise subset. For example, when the first text in the premise includes a hypothetical word, the premise can be added to the third premise subset. For example, when the similarity between the first text and the second text in the premise is lower than the similarity threshold, the premise can be added to the third premise subset.

[0126] In step S23, a second premise subset is obtained according to a score corresponding to any determined premise in the first premise subset;

[0127] The specific process is as described above and will not be repeated here.

[0128] The first premise subset includes at least one determined premise in the premise set.

[0129] According to some embodiments, obtaining the second premise subset according to a score corresponding to any determined premise in the first premise subset includes:

[0130] determining a target determination premise according to a first score of any premise in the first premise subset;

[0131] According to the second score of any premise in the premise subset except the target-determined premise, selecting at least one premise that meets the score requirement;

[0132] The target determination prerequisite and at least one prerequisite that satisfies the score requirement are added to the second prerequisite subset.

[0133] According to some embodiments, determining a target determination premise according to a first score of any premise in the first premise subset includes:

[0134] Determine a first score corresponding to any determined premise according to the correlation and logical relationship between any determined premise in the first premise subset and the target hypothesis;

[0135] According to the first score corresponding to any of the determined premises, the determined premise with the highest first score in the first premise subset is taken as the target determined premise;

[0136] According to the second score of any prerequisite other than the target-determined prerequisite in the prerequisite subset, before selecting at least one prerequisite that meets the score requirement, the method further includes:

[0137] According to the correlation and logical relationship between any premise in the premise set except the target determination premise and the target determination premise, the second score corresponding to any premise is determined. Therefore, the non-determined premise can be generated into a deterministic premise according to the score, which can reduce the difficulty of determining the reasoning result and improve the efficiency and accuracy of determining the reasoning result. In addition, the reasoning priority can be adjusted based on the score, which can improve the reasoning efficiency.

[0138] According to some embodiments, the first score may be determined, for example, by using a scoring function in a prompt module of a large model. The scoring function may, for example, indicate the interaction score of each part, and may also indicate the interaction score of each sentence structure.

[0139] Among them, in one embodiment of the present disclosure, the score requirement refers to the requirement for determining the second premise subset. The score requirement does not specifically refer to a fixed requirement. For example, the score requirement may be a requirement greater than a score threshold, and the score requirement may also be, for example, selecting a preset number of premises in descending order of the second scores. The present disclosure embodiment is not limited by comparison.

[0140] In step S24, the second premise subset is merged and inferred to obtain the inference premise corresponding to the second premise subset;

[0141] According to some embodiments, when the second premise subset is obtained, the second premise subset can be merged and inferred to obtain the inference premise corresponding to the second premise subset. For example, the second premise subset can be inferred as a whole to obtain the inference premise corresponding to the second premise subset. The inference premise can be, for example, a different inference premise from any premise included in the premise subset. For example, before the second premise subset is inferred, the premise included in the second premise subset can be merged. The inference premise is to determine the priority.

[0142] In step S25, at least one verification operation is performed on the inference premise, and when the verification results corresponding to the at least one verification operation are all verified, it is determined that the inference premise that has passed the verification is obtained;

[0143] The specific process is as described above and will not be repeated here.

[0144] According to some embodiments, performing at least one verification operation on the inference premise includes at least one of the following:

[0145] Logically verify the premises of reasoning;

[0146] Conduct source checking and verification of the premises of reasoning;

[0147] Repeated verification of the premise of reasoning;

[0148] Verify the validity of the premise of reasoning;

[0149] Verify the boundary conditions of the reasoning premise. Therefore, verifying the reasoning premise can improve the accuracy of obtaining the reasoning premise and the accuracy of determining the reasoning result.

[0150] The logic verification may be, for example, a global verification, which is used to check whether the inference premise complies with the logic of all current premises.

[0151] The verification may also include local verification, for example. The local verification may be used to verify whether the inference premise is obtained from the screened premise set.

[0152] The source verification may be used, for example, to verify whether unknown source information other than the provided premise set is used.

[0153] The boundary condition verification may be used, for example, to check whether a boundary condition is violated.

[0154] Among them, validity verification can be used, for example, to check whether it is helpful in obtaining the reasoning results of the target hypothesis.

[0155] Among them, repeatability verification can be used, for example, to check whether the inference premise is only a part of any premise in a certain premise set.

[0156] For example, format checking may also be included, wherein the format checking may be used to check whether the format of the inference premise meets the format requirements, or whether the large model gives a negative answer.

[0157] In step S26, the verified inference premise is added to the premise set;

[0158] The specific process is as described above and will not be repeated here.

[0159] According to some embodiments, the method further comprises:

[0160] If any of the verification results corresponding to at least one verification operation fails, the verification failure reasons corresponding to the second premise subset and the reasoning premise are recorded. Therefore, in the next iteration operation, the repeated verification of the reasoning premise that failed the verification can be reduced, the determination time of the reasoning result can be reduced, and the efficiency of determining the reasoning result can be improved.

[0161] In step S27, when the iterative addition operation of the premise set meets the preset iteration requirements, a question operation is performed on the premise set and the target hypothesis after the iteration to obtain the question operation result;

[0162] The specific process is as described above and will not be repeated here.

[0163] According to some embodiments, when the iterative addition operation of the premise set meets the preset iteration requirement, a questioning operation is performed on the premise set and the target hypothesis after the iteration to obtain the questioning operation result, including:

[0164] When the number of verification failures corresponding to the premise set reaches a number threshold, a question operation is performed on the premise set after iteration and the target hypothesis to obtain a question operation result; or

[0165] When the number of inference premises corresponding to the premise set reaches the quantity threshold, the premise set and the target hypothesis after the iteration are questioned to obtain the questioning operation result. Therefore, the iteration operation can be stopped when the preset iteration requirements are met, which can improve the accuracy of the questioning operation result determination.

[0166] The number threshold is used to indicate the threshold for stopping the iteration operation according to the number of verification failures, and the number threshold does not specifically refer to a fixed threshold. For example, when a modification instruction for the number threshold is received, the number threshold may also change accordingly. For example, when the specific number of times corresponding to the number threshold changes, the number threshold may also change accordingly.

[0167] In some embodiments, the questioning operation may be performed by, for example, asking multiple times using a chain of thought, and after the results are obtained, a chain of thought (CoT-SC) is selected, and the result of the questioning operation is determined by voting.

[0168] Among them, Chain of Thought (CoT), for example, can be an improved prompting strategy for improving the performance of LLM in complex reasoning tasks, such as arithmetic reasoning, commonsense reasoning, and symbolic reasoning.

[0169] In some embodiments, for example, it can be determined whether the reasoning is finished based on the entire history of the reasoning. When it is determined to be finished, a question operation can be performed on the iterated premise set and the target hypothesis to obtain the question operation result.

[0170] In step S28, the inference result of the target question is determined according to the question operation result;

[0171] The specific process is as described above and will not be repeated here.

[0172] In some embodiments, the inference result of the target question is determined based on the question operation result. For example, the number of times the same question operation result is obtained in the question operation results, and the question operation result with the largest number is used as the inference result of the target question.

[0173] In step S29, when the reasoning result of the target question is not determined according to the question operation result, supplementary reasoning is performed on the target question according to the reasoning process corresponding to the target question to obtain the supplementary reasoning result corresponding to the target question.

[0174] In some embodiments, when the question operation result is obtained and the reasoning result of the target question is not determined, supplementary reasoning is performed on the target question according to the reasoning process corresponding to the reasoning result to obtain the supplementary reasoning result corresponding to the target question.

[0175] The reasoning process includes at least one of an iteration process, a verification process, and a questioning process. For example, according to the entire reasoning failure process corresponding to the target problem, the target is reasoned and supplemented to obtain a supplementary reasoning result of the target problem.

[0176] For example, the entire reasoning can be completed, for example, the reasoning steps from the current state to the target hypothesis can be given. If the reasoning steps do not meet the reasoning conditions, the reasoning can be repeated. The reasoning steps do not meet the reasoning conditions, for example, there are multiple correct options in the reasoning steps, there are no correct options in the reasoning steps, there are obvious contradictions between the multiple premises of the deduction, etc.

[0177] According to some embodiments, Tables 1 to 4 show reasoning information when the technical solution of the present disclosure is not adopted, and Table 5 shows reasoning information when the technical solution of the present disclosure is adopted.

[0178] Table 1

[0179]

[0180] Table 2

[0181]

[0182] Table 3

[0183]

[0184] Table 4

[0185]

[0186] Among them, the accuracy rate can be, for example, the number of questions correctly answered by the model / the total number of questions; the average number of state visits is the number of state visits required for the model to solve a problem. In ToT, one state corresponds to a node expansion in the solution space, and in CR and DetermLR, one state corresponds to a premise set acquisition and an inference process (success\failure).

[0187] Table 5

[0188]

[0189] Based on Table 5, it can be seen that the number of state visits required by the technical solution disclosed in the present invention is lower than that of other solutions, which can improve the accuracy while improving the reasoning efficiency.

[0190] In some or related embodiments, a large language model is used to classify the premise set, and a first premise subset and a third premise subset are obtained. The second premise subset can be obtained according to the score corresponding to any definite premise in the first premise subset, and the premise in the premise set that is most relevant to the target hypothesis can be obtained. The second premise subset can be determined based on the remaining premises and the most relevant premise. Since the remaining premises include uncertain premises, and the inference premise can be generated by inference based on the second premise subset, the inference premise is a definite premise, so the transformation mechanism from uncertain premise to definite premise can be adjusted, the inference structure can be adjusted to match the inference task corresponding to the target problem, reduce the complexity of the inference task, improve the inference efficiency, and determine the priority and inference order through the score, improve the new inference path, and improve the inference efficiency. In addition, the second premise subset is merged and inferred to obtain the inference premise corresponding to the second premise subset; at least one verification operation is performed on the inference premise, and when the verification results corresponding to at least one verification operation are all verified, it is determined that the inference premise that has passed the verification is obtained; the inference premise is added to the premise set, so that the accuracy of the determination of the inference premise can be improved. When the iterative addition operation of the premise set meets the preset iteration requirements, a question operation is performed on the premise set and the target hypothesis after the iteration to obtain the question operation result; based on the question operation result, the reasoning result of the target problem is determined; and questioning can be used to provide reasoning experience to improve the accuracy of determining the reasoning result.

[0191] Figure 4 is a block diagram of a device for determining an inference result according to an exemplary embodiment. Figure 4 , the device comprises:

[0192] The problem acquisition unit 401 is used to acquire a target problem, wherein the target problem includes a premise set and a target hypothesis, the premise set includes at least one premise, the premise includes condition information corresponding to the target problem, and the target hypothesis is a hypothesis to be determined proposed for the premise set;

[0193] The set acquisition unit 402 is used to acquire a second premise subset according to a score corresponding to any determined premise in the first premise subset, wherein the first premise subset includes at least one determined premise in the premise set, and the second premise subset includes at least one determined premise and / or at least one undetermined premise in the premise set, the determined premise is related to the target hypothesis and does not carry condition information in a selection format, and the undetermined premise is unrelated to the target hypothesis or carries condition information in a selection format;

[0194] The inference premise adding unit 403 is used to obtain the verified inference premise corresponding to the second premise subset, and add the inference premise to the premise set, wherein the inference premise is the premise obtained by inferring the second premise subset;

[0195] A result acquisition unit 404 is used to perform a questioning operation on the premise set and the target hypothesis after the iteration to obtain the questioning operation result when the iterative addition operation of the premise set meets the preset iteration requirement;

[0196] The result determination unit 405 is used to determine the inference result of the target question according to the question operation result.

[0197] According to some embodiments, the set acquisition unit 402 is used to acquire the second premise subset according to the score corresponding to any determined premise in the first premise subset, specifically to:

[0198] determining a target determination premise according to a first score of any premise in the first premise subset;

[0199] According to the second score of any premise in the premise subset except the target-determined premise, selecting at least one premise that meets the score requirement;

[0200] The target determination prerequisite and at least one prerequisite that satisfies the score requirement are added to the second prerequisite subset.

[0201] According to some embodiments, the set acquisition unit 402 is configured to determine the target determination premise according to the first score of any premise in the first premise subset, specifically to:

[0202] Determine a first score corresponding to any determined premise according to the correlation and logical relationship between any determined premise in the first premise subset and the target hypothesis;

[0203] According to the first score corresponding to any of the determined premises, the determined premise with the highest first score in the first premise subset is taken as the target determined premise;

[0204] According to the degree of association and logical relationship between any premise in the premise set except the target determination premise and the target determination premise, a second score corresponding to any premise is determined.

[0205] According to some embodiments, the set acquisition unit 402 is further configured to, before acquiring the second premise subset according to a score corresponding to any determined premise in the first premise subset:

[0206] The premise set is classified by using a large language model to obtain a first premise subset and a second premise subset, wherein the first premise subset includes at least one definite premise and the third premise subset includes at least one indefinite premise.

[0207] According to some embodiments, the reasoning premise adding unit 403 is used to obtain the verified reasoning premise corresponding to the second premise subset, specifically to:

[0208] Merging and reasoning the second premise subset to obtain the reasoning premise corresponding to the second premise subset;

[0209] At least one verification operation is performed on the inference premise, and when the verification results corresponding to the at least one verification operation are all verified to be passed, it is determined that the verified inference premise is obtained.

[0210] According to some embodiments, the inference premise adding unit 403 is further configured to:

[0211] In the case that any verification result corresponding to at least one verification operation fails the verification, the verification failure reason corresponding to the second premise subset and the reasoning premise is recorded.

[0212] According to some embodiments, the reasoning premise adding unit 403 is used to perform at least one verification operation on the reasoning premise, specifically for at least one of the following:

[0213] Logically verify the premises of reasoning;

[0214] Conduct source checking and verification of the premises of reasoning;

[0215] Repeated verification of the premise of reasoning;

[0216] Verify the validity of the premise of reasoning;

[0217] Verify the boundary conditions of the reasoning premise.

[0218] According to some embodiments, the reasoning premise adding unit 403 is used to perform a questioning operation on the premise set and the target hypothesis after the iteration when the iterative addition operation of the premise set meets the preset iteration requirement, and when obtaining the questioning operation result, is specifically used to:

[0219] When the number of verification failures corresponding to the premise set reaches a number threshold, a question operation is performed on the premise set after iteration and the target hypothesis to obtain a question operation result; or

[0220] When the number of reasoning premises corresponding to the premise set reaches a quantity threshold, a question operation is performed on the iterated premise set and the target hypothesis to obtain a question operation result.

[0221] According to some embodiments, the result determination unit 405 is further configured to:

[0222] When the reasoning result of the target question is not determined according to the questioning operation result, supplementary reasoning is performed on the reasoning result according to the reasoning process corresponding to the reasoning result and the target question, wherein the reasoning process includes at least one of an iterative process, a verification process and a questioning process.

[0223] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0224] In some or related embodiments, a target problem is obtained by a problem acquisition unit, wherein the target problem includes a premise set and a target hypothesis; a set acquisition unit is used to obtain a second premise subset according to a score corresponding to any determined premise in the first premise subset, wherein the first premise subset includes at least one determined premise in the premise set; a reasoning premise adding unit is used to obtain a verified reasoning premise corresponding to the second premise subset, and add the reasoning premise to the premise set; a result acquisition unit is used to perform a question operation on the premise set and the target hypothesis after the iteration when the iterative addition operation of the premise set meets the preset iteration requirements, and obtain the question operation result; a result determination unit is used to determine the reasoning result of the target problem according to the question operation result. Therefore, without using external tools, the reasoning structure corresponding to the target problem can be obtained, and the reasoning structure can be adjusted by reasoning according to the determined premise, so as to improve the matching of the reasoning result and the reasoning process, reduce the mismatch between the reasoning result and the reasoning process, and make the reasoning process complicated, and the premise can be quantified by scores to improve the iteration efficiency, so that the reasoning efficiency and reasoning accuracy can be improved without increasing the reasoning cost.

[0225] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable electronic devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0226] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0227] Multiple components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the electronic device 500 to exchange information / data with other electronic devices through a computer network such as the Internet and / or various telecommunication networks.

[0228] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the reasoning result determination method. For example, in some embodiments, the reasoning result determination method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the reasoning result determination method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the reasoning result determination method in any other appropriate manner (e.g., by means of firmware).

[0229] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0230] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0231] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0232] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0233] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0234] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0235] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0236] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining an inference result, characterized in that: include: Acquire a target problem, wherein the target problem includes a premise set and a target hypothesis, the premise set includes at least one premise, the premise includes condition information corresponding to the target problem, and the target hypothesis is a hypothesis to be determined proposed for the premise set; According to the score corresponding to any determined premise in the first premise subset, a second premise subset is obtained, wherein the first premise subset includes at least one determined premise in the premise set, and the second premise subset includes at least one determined premise and / or at least one undetermined premise in the premise set, the determined premise is condition information related to the target hypothesis and does not carry a selection format, and the undetermined premise is condition information that is unrelated to the target hypothesis or carries a selection format; Acquire a verified inference premise corresponding to the second premise subset, and add the inference premise to the premise set, wherein the inference premise is a premise obtained by inferring the second premise subset; When the iterative addition operation of the premise set meets the preset iteration requirement, performing a questioning operation on the premise set after the iteration and the target hypothesis to obtain a questioning operation result; According to the question operation result, the inference result of the target question is determined.

2. The method according to claim 1, characterized in that The step of obtaining the second premise subset according to a score corresponding to any determined premise in the first premise subset includes: Determining a target determination premise according to a first score of any premise in the first premise subset; selecting at least one premise that meets the score requirement according to a second score of any premise in the premise subset except the target-determined premise; The target determination prerequisite and the at least one prerequisite that satisfies the score requirement are added to a second subset of prerequisites.

3. The method according to claim 2, characterized in that Determining a target determination premise according to a first score of any premise in the first premise subset includes: Determine a first score corresponding to any determined premise in the first premise subset according to the degree of association and the logical relationship between the target hypothesis and the determined premise; According to the first score corresponding to any of the determined premises, taking the determined premise with the highest first score in the first premise subset as the target determined premise; Before selecting at least one premise that meets the score requirement according to the second score of any premise in the premise subset except the target-determined premise, the method further includes: According to the degree of association and the logical relationship between any premise in the premise set except the target determination premise and the target determination premise, a second score corresponding to the any premise is determined.

4. The method according to any one of claims 1 to 3, characterized in that Before obtaining the second premise subset according to the score corresponding to any determined premise in the first premise subset, the method further includes: The premise set is classified by using a large language model to obtain a first premise subset and a second premise subset, wherein the first premise subset includes at least one definite premise, and the third premise subset includes at least one indefinite premise.

5. The method according to claim 1, characterized in that The obtaining of the verified reasoning premise corresponding to the second premise subset includes: Merging and reasoning the second premise subset to obtain reasoning premises corresponding to the second premise subset; At least one verification operation is performed on the inference premise, and when the verification results corresponding to the at least one verification operation are all verified to be passed, it is determined that the verified inference premise is obtained.

6. The method according to claim 4, characterized in that The method further comprises: In the event that any verification result corresponding to the at least one verification operation fails the verification, the verification failure reason corresponding to the second premise subset and the reasoning premise is recorded.

7. The method according to claim 4, characterized in that The performing at least one verification operation on the reasoning premise includes at least one of the following: Logically verify the premise of the reasoning; Conduct source checking and verification of the stated reasoning premises; Repeatable verification of the reasoning premise; Verifying the validity of the reasoning premise; The boundary conditions of the reasoning premise are verified.

8. The method according to claim 7, characterized in that When the iterative addition operation of the premise set meets the preset iteration requirement, a questioning operation is performed on the premise set after the iteration and the target hypothesis to obtain a questioning operation result, including: When the number of verification failures corresponding to the premise set reaches a number threshold, performing a questioning operation on the iterated premise set and the target hypothesis to obtain a questioning operation result; or When the number of inference premises corresponding to the premise set reaches a quantity threshold, a question operation is performed on the iterated premise set and the target hypothesis to obtain a question operation result.

9. The method according to claim 1, characterized in that: The method further comprises: When the reasoning result of the target question is not determined according to the questioning operation result, supplementary reasoning is performed on the target question according to the reasoning process corresponding to the target question to obtain the supplementary reasoning result corresponding to the target question, wherein the reasoning process includes at least one of an iterative process, a verification process and a questioning process.

10. A device for determining an inference result, characterized in that: include: A question acquisition unit, which acquires a target question, wherein the target question includes a premise set and a target hypothesis; a set acquisition unit, configured to acquire a second premise subset according to a score corresponding to any determined premise in the first premise subset, wherein the first premise subset includes at least one determined premise in the premise set, the premise set includes at least one premise, the premise includes condition information corresponding to the target problem, and the target hypothesis is a hypothesis to be determined proposed for the premise set; an inference premise adding unit, configured to obtain a verified inference premise corresponding to the second premise subset, and add the inference premise to the premise set, wherein the second premise subset includes at least one determined premise and / or at least one undetermined premise in the premise set, wherein the determined premise is condition information related to the target hypothesis and does not carry a selection format, and the undetermined premise is condition information that is unrelated to the target hypothesis or carries a selection format; A result acquisition unit is used to perform a questioning operation on the premise set after the iteration and the target hypothesis when the iterative addition operation of the premise set meets the preset iteration requirement, and obtain a questioning operation result, wherein the inference premise is a premise obtained by inferring the second premise subset; The result determination unit is used to determine the inference result of the target question according to the question operation result.

11. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the inference result determination method as described in any one of claims 1 to 9.

12. A storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to execute the inference result determination method as claimed in any one of claims 1 to 9.

Images