arXiv Computation and Language

Reasoning Error from Known Fact: Step-Level Self-Consistency Group Relative Policy Optimization for LLMs

The paper introduces Step-level Self-Consistency Group Relative Policy Optimization (SSC‑GRPO), a method that assigns rewards to individual steps in a reasoning trace by evaluating self‑consistency across multiple rollouts. It targets Context‑Sensitive Factual Hallucinations—errors that arise when a model has the correct knowledge but produces false statements due to contextual interference. SSC‑GRPO outperforms previous approaches on mathematical reasoning benchmarks and hallucination leaderboards, offering a new way to detect and reduce hallucinations in large language model reasoning.

arXiv Computation and Language

Aug 25

GRACE: Step-Level Benchmark for Faithful Reasoning over Context

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

llms nlp reinforcement-learning benchmarks safety

arXiv AI

Jun 15

ClinHallu: A Benchmark for Diagnosing Stage-Wise Hallucinations in Medical MLLM Reasoning

arXiv:2606. 14697v1 Announce Type: cross Abstract: Building trustworthy medical multimodal large language models (MLLMs) is critical for reliable clinical decision support.

By Sicheng Yang, Hangjie Yuan, Wenjun Zhang, Jinwang Wang, Yichen Qian, Weihua Chen, Fan Wang, Lei Zhu

llms fine-tuning multimodal benchmarks safety

arXiv AI

Jul 21

Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation Steering

arXiv:2607. 18100v1 Announce Type: new Abstract: Extended reasoning has become standard for frontier Large Language Models (LLMs), yet the trajectories these models produce remain largely uncontrollable.

By Sheldon Yu, Tong Yu, Xunyi Jiang, Rohan Surana, Gagan Mundada, Sungchul Kim, Lina Yao, Julian McAuley, Junda Wu

llms

arXiv Machine Learning

Jun 4

Reasoning Shift: How Context Silently Shortens LLM Reasoning

arXiv:2604. 01161v2 Announce Type: replace Abstract: Large language models (LLMs) exhibiting test-time scaling behavior, such as extended reasoning traces and self-verification, have demonstrated remarkable performance on complex, long-term reasoning tasks.

By Gleb Rodionov, Roman Garipov, George Yakushev

llms agents fine-tuning

arXiv Machine Learning

Jul 27

Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning Models

arXiv:2607. 22098v1 Announce Type: cross Abstract: Large reasoning models (LRMs) generate long reasoning traces before producing final answers.

By Junlin Fang, Do Nguyen-Thanh, Xiaogang Xu, Zhen Fang, Sean Du

rag benchmarks safety

arXiv AI

Sep 21

LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers

arXiv:2609.21492v1 Announce Type: new Abstract: Chain-of-Thought (CoT) reasoning has been shown to improve the performance of large language models (LLMs), yet existing optimization methods largely r...

By Jingyu Hu, Shu Yang, Weiru Liu, Di Wang

llms fine-tuning benchmarks

arXiv Computation and Language

Oct 8

PHRBench: A Behavioral Evaluation of Post-Hallucination Reasoning in LLMs

PHRBench is a new benchmark that evaluates how large language models handle hallucinated information during multi-stage reasoning. It tests 18 models across four domains using 4,820 controlled instances, measuring Hallucination Compliance, Hallucination Avoidance, and Heuristic Correction to determine if a model can correct hallucinated premises and still reach the correct answer. The study finds that successful recovery is rare, often requires frequent belief updates, and that features of the hallucinated prompt can predict recovery with an AUROC of 0.847.

By Linghao Meng, Feng He, Xuan Yang, Junyuan Mao, Pinze Ren, Deqing Mu, Hesen Yang, Qiankun Li

llms benchmarks safety

arXiv AI

Jul 9

HIVE: Understanding Post-Hallucination Reasoning in Vision Language Models

arXiv:2607. 07507v1 Announce Type: cross Abstract: Hallucinations in vision language models (VLMs) are commonly treated as semantic errors, yet they often arise from partial or ambiguous visual evidence.

By Feng He, Zhenting Wang, Qifan Wang, Qiang Guan, Dongfang Liu, Ruixiang Tang, Qiankun Li

llms multimodal safety

arXiv AI

Jun 9

Diagnosing Multi-step Reasoning Failures in Black-box LLMs via Stepwise Confidence Attribution

arXiv:2605. 19228v2 Announce Type: replace-cross Abstract: Large Language Models have achieved strong performance on reasoning tasks with objective answers by generating step-by-step solutions, but diagnosing where a multi-step reasoning trace might fail remains difficult.

By Xiaoou Liu, Tiejin Chen, Dengjia Zhang, Yaqing Wang, Lu Cheng, Hua Wei

llms nlp

arXiv AI

Jun 6

ReasoningFlow: Discourse Structures for Understanding LLM Reasoning Traces

arXiv:2606. 05402v1 Announce Type: cross Abstract: Large reasoning models (LRMs) produce reasoning traces with non-linear structures, such as backtracking and self-correction, that complicate the evaluation and monitoring of the reasoning process.

By Jinu Lee, Shivam Agarwal, Amruta Parulekar, Siddarth Madala, Dilek Hakkani-Tur, Julia Hockenmaier

llms

arXiv Machine Learning

Jun 2

FLaG: Fine-Grained Latent Grouping for Hallucination Detection

arXiv:2606. 00301v1 Announce Type: new Abstract: Hallucinations in large language models (LLMs) arise from heterogeneous failure mechanisms, making reliable detection difficult for any single global uncertainty score.

By Wentao Ye, Liyao Li, Zhiqing Xiao, Muzhi Zhu, Jiaqi Hu, Zhanming Shen, Xiaomeng Hu, Sean Du, Haobo Wang

llms benchmarks safety

arXiv AI

Jul 28

Reason Popper-ly: Patching In-Context Reasoning with Inductive Logic Programming

arXiv:2607. 23019v1 Announce Type: new Abstract: Chain-of-thought (CoT) prompting enables large language models (LLMs) to tackle multi-step reasoning tasks, yet the generated intermediate steps are not guaranteed to be logically sound.

By Zirong Chen, Meiyi Ma

llms benchmarks

Related stories