# GRACE: Step-Level Benchmark for Faithful Reasoning over Context

arXiv: https://arxiv.org/abs/2606.16151 · HTML: https://mlsanity.com/paper/2606.16151/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 49 references that are in the map.
- First posted on arXiv 15 Jun 2026.

## Builds on (most cited first, 23 of 23)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation](https://mlsanity.com/paper/2305.01210/) (2023)
- [Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps](https://mlsanity.com/paper/2011.01060/) (2020)
- [FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://mlsanity.com/paper/2305.14251/) (2023)
- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations](https://mlsanity.com/paper/2312.08935/) (2023)
- [Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions](https://mlsanity.com/paper/2212.10509/) (2022)
- [Improve Mathematical Reasoning in Language Models by Automated Process Supervision](https://mlsanity.com/paper/2406.06592/) (2024)
- [A Survey of Hallucination in Large Foundation Models](https://mlsanity.com/paper/2309.05922/) (2023)
- [RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models](https://mlsanity.com/paper/2401.00396/) (2023)
- [ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systems](https://mlsanity.com/paper/2311.09476/) (2023)
- [OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement](https://mlsanity.com/paper/2402.14658/) (2024)
- [ReClor: A Reading Comprehension Dataset Requiring Logical Reasoning](https://mlsanity.com/paper/2002.04326/) (2020)
- [ROSCOE: A Suite of Metrics for Scoring Step-by-Step Reasoning](https://mlsanity.com/paper/2212.07919/) (2022)
- [Fine-grained Hallucination Detection and Editing for Language Models](https://mlsanity.com/paper/2401.06855/) (2024)
- [A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chains](https://mlsanity.com/paper/2402.00559/) (2024)
- [VersaPRM: Multi-Domain Process Reward Model via Synthetic Reasoning Data](https://mlsanity.com/paper/2502.06737/) (2025)
- [Understanding Chain-of-Thought in LLMs through Information Theory](https://mlsanity.com/paper/2411.11984/) (2024)
- [FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning](https://mlsanity.com/paper/2510.04040/) (2025)

## Built on by (most cited first, 2 of 2)

- [Two Regimes of Chain-of-Thought Unfaithfulness: Metric-Based Detection Fails Where Models Are Wrong](https://mlsanity.com/paper/2607.23458/) (2026)
- [Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs](https://mlsanity.com/paper/2608.11573/) (2026)

## Abstract (arXiv)

Many reasoning tasks require models to reason over input context, from document-grounded question answering to rule-based deduction. Chain-of-Thought (CoT) prompting produces traces that appear transparent, yet individual steps can silently deviate from the source evidence, even when the final answer is correct. Existing methods detect hallucinations at the response level but fail to identify where in the chain a failure occurs or what type it is. We introduce GRACE, the first human-annotated step-level faithfulness benchmark with a data-driven error taxonomy for context-grounded textual reasoning. GRACE covers CoT traces from 10 models across 4 source datasets, with each step annotated for faithfulness, error category, and natural language explanation. A data-driven taxonomy, discovered bottom-up via unsupervised clustering, organizes failures into two tracks: GRACE-Inference (deductive errors) and GRACE-Grounding (factual grounding errors), with four categories each. The evaluation set is human-annotated and challenging by design. Our experiments reveal substantial headroom for current models. In addition, integrating step-level faithfulness signals into reinforcement learning pipelines improves both downstream accuracy and reasoning reliability.