GRACE is a step‑level benchmark for evaluating the faithfulness of chain‑of‑thought reasoning over context. It provides human annotations for each step in CoT traces from 10 models across 4 datasets, labeling faithfulness, error category, and natural‑language explanations. The benchmark introduces a data‑driven taxonomy that splits errors into GRACE‑Inference (deductive) and GRACE‑Grounding (factual) tracks, each with four categories, and demonstrates that incorporating step‑level faithfulness signals can improve downstream accuracy and reasoning reliability.
By Hoang Pham, Dong Le, Anh Tuan Luu
The paper introduces Fact-Ablated Evaluation (FAE), a framework that iteratively removes cited evidence to test whether large language models (LLMs) adjust their fact‑checking predictions accordingly. Experiments reveal that many off‑the‑shelf LLMs rely more on internal knowledge than on the provided evidence. To address this, the authors propose REAL, a training method that uses counterfactual evidence supervision to encourage LLMs to base veracity judgments on evidence, achieving better evidence‑dependent performance across four datasets.
By Xingyu Deng, Mingzi Cao, Nikolaos Aletras, Xi Wang, Mark Stevenson
Large language models increasingly rely on long-form reasoning for complex tasks, yet their reasoning traces may drift away from the supplied context when evidence is sparse, noisy, or in conflict with parametric knowledge. Existing grounding methods either attach citations after generation or encourage evidence retrieval inside the trace, but they often do not ensure that cited content is sufficient for the local inference and final answer.
arXiv:2604. 13201v2 Announce Type: replace-cross Abstract: Large language models are emerging as scientific assistants, but evaluating their ability to reason from empirical data remains challenging.
By Oliver Bentham, Vivek Srikumar
The paper introduces LURCH, a benchmark of 6,550 question‑answering instances that quantifies long‑context reasoning difficulty using three properties: evidence scope, evidence dispersion, and reasoning depth. Experiments with both proprietary and open‑weight models show that these properties are strongly negatively correlated with performance, that performance drops sharply for dispersed evidence as context length grows, and that difficulty increases when multiple properties are high simultaneously. The study demonstrates that context length alone is insufficient to explain model failures, and that the new metrics provide a more nuanced understanding of long‑context challenges.
By Shiwen Qin, Cameron Barker, Henry Gouk, Elliot J. Crowley, Shay B. Cohen, Mirella Lapata, Miao Li
Understanding and reasoning over long contexts has become a key requirement for deploying large language models (LLMs) in realistic applications. Although recent LLMs support increasingly long context windows, they often fail to use relevant evidence that is already present in the input, revealing a gap between context access and effective context utilization.