[KongchangAI]
Model

ChronoRAG-G

A research system based on the OETRA method for time-sensitive evidence retrieval and answer generation, capable of preserving source identity and temporal information within graph-structured contextual corpora.

Timeline (last 90 days)

Nov 9

The author built the research system ChronoRAG-G, whose underlying method is OETRA (Obligation-Explicit Temporal Retrieval and Allocation).

Unverified50%
Nov 9

The author evaluated on 480 updated earnings call transcripts using 1,005 temporal QA questions.

Unverified50%
Nov 9

The evaluation results show that ChronoRAG-G achieves an 80.70% audited answer accuracy.

Unverified50%
Nov 9

The evaluation results show that ChronoRAG-G achieves a 95.92% strict traceable slot coverage rate.

Unverified50%
Nov 9

In the evaluation, 167 answers judged incorrect had complete evidence traceability chains, showing that successful evidence parsing does not guarantee a correct final answer.

Unverified50%
Nov 9

The author explicitly states that the above application directions are potential directions rather than claims of production deployment, and the current evaluation is limited to financial temporal QA.

Unverified50%

All Facts (6)

Source Articles