Decision-level state context snapshot generation, binding, and playback system

Abstract

A system for generating, binding, and replaying state context snapshots at the decision level includes a transmission interface, memory, and a processor electrically connected to the transmission interface and memory. The system dynamically assembles multiple pieces of context data according to temporal semantic rules and the corresponding temporal information for each context data to generate structured data. It then performs dynamic lexical pruning on the structured data based on lexical budgeting to generate state context snapshots. Finally, it performs hash operations on the state context snapshots to establish a correlation and binding with the decision output results, and serializes the state context snapshots for subsequent reloading to reconstruct the state context at the decision point. This achieves the technical effect of improving the reconstructability and verifiability of the state context at the decision point in artificial intelligence.

Claims

1. A system for generating, binding, and replaying state context snapshots at a decision level, the system comprising: a transmission interface for receiving multiple pieces of contextual data, wherein each piece of contextual data is associated with a decision point in time and corresponds to temporal information; a memory for storing multiple computer-executable instructions; and a processor electrically connected to the transmission interface and the memory, the processor executing the multiple computer-executable instructions to be configured to: dynamically assemble the multiple pieces of contextual data from the transmission interface according to a temporal semantic rule and the temporal information corresponding to each piece of contextual data to generate structured data corresponding to the decision point in time; perform a dynamic lexical pruning process on the structured data according to a lexical budget to control the amount of structured data within the lexical budget to generate a state context snapshot; perform a hash operation on the state context snapshot to generate a snapshot hash value corresponding to the state context snapshot; and use the state context snapshot as a decision input to an artificial intelligence model, causing the artificial intelligence model to generate a decision output result based on the state context snapshot. Establish and record the association between the snapshot hash value and one of the decision output results; and serialize the state context snapshot to store it as serialized data, and subsequently reload the serialized data to reconstruct the state context at one of the decision points.

2. A system for generating, binding, and replaying state context snapshots of decision levels, as requested in item 1, wherein the multiple context data include multi-dimensional attribute data such as user persona, role, paragraph, risk level, and policy version.

3. The system for generating, binding, and replaying state context snapshots at the decision level as requested in Request 1, wherein the processor is further configured to: establish a version identification information and a state change record for the state context snapshot to form a traceable versioned context data; and set a time validity period for the state context snapshot, and restrict the use of the state context snapshot by subsequent decision-making procedures when a system time exceeds the time validity period.

4. A system for generating, binding, and replaying state context snapshots at the decision level, as requested in item 1, wherein the processor executes a multi-dimensional state context combination algorithm to process multiple data dimensions to generate the structured data, the multi-dimensional state context combination algorithm comprising: Based on the time information corresponding to each of the aforementioned situational data, select multiple valid situational data that are still valid at the decision time from the multiple situational data; map each of the multiple valid situational data to a field of the corresponding data dimension in a predetermined multi-dimensional structure; and according to the predetermined multi-dimensional structure, integrate the corresponding multiple valid situational data into structured data having the field corresponding to each of the data dimensions.

5. The system for generating, binding, and replaying state context snapshots at the decision level as described in Request 1, wherein when the processor establishes and records the association binding between the snapshot hash value and the decision output result, the processor is further configured to: write the snapshot hash value, the decision output result, and binding time information into a binding record; and in a verification procedure, re-execute the hash operation on the state context snapshot to obtain a verification hash value, and compare the verification hash value with the snapshot hash value in the binding record, so as to determine that at least one of the state context snapshot and the decision output result has been tampered with if the two are inconsistent.

Images