# ContractEval: Improves Procedural Instruction Conformance

Digest AI · Generative AI & Models · published 2026-09-11T04:00:00Z

Canonical: https://digestai.news/story/contracteval-improves-procedural-instruction-conformance

## Summary

As LLM agents move from answering questions to executing procedures, failures can be subtle. The article introduces **ContractEval**, a diagnostic framework that makes the active obligations in procedural instructions explicit. It matches query-active obligations against response or trace evidence to identify and localize structural failures such as omissions, wrong branches, ordering errors, extra actions, invariant breaches, and output-contract violations. On a suite of audited procedural contracts, **ContractEval** outperforms both output-only and trace-aware LLM judges by detecting all the injected structural failures. The framework is not a guarantee but makes conformance auditing more transparent.

## Key points

- Introduces ContractEval for procedural instruction conformance
- Matches query-active obligations against response or trace evidence
- Detects and localizes structural failures in procedures

## Why it matters

Improves the reliability of LLM agents executing complex procedures by making their actions more transparent.

## Sources

1. [ContractEval: Query-Conditioned Execution Matching for Procedural Instruction Conformance](https://arxiv.org/abs/2609.09458) (arXiv cs.AI, 2026-09-11, primary source)

## Cite

Digest AI, "ContractEval: Improves Procedural Instruction Conformance", 11 September 2026, https://digestai.news/story/contracteval-improves-procedural-instruction-conformance

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/contracteval-improves-procedural-instruction-conformance.json