From Harm to Help: Turning Reasoning In-Context Demos into Assets for Reasoning LMs

• URL: http://arxiv.org/abs/2509.23196v1
• Date: Sat, 27 Sep 2025 08:59:31 GMT
• Title: From Harm to Help: Turning Reasoning In-Context Demos into Assets for Reasoning LMs
• Authors: Haonan Wang, Weida Liang, Zihang Fu, Nie Zheng, Yifan Zhang, Yao Tong, Tongyao Zhu, Hao Jiang, Chuang Li, Jiaying Wu, Kenji Kawaguchi,
• Abstract summary: We revisit this paradox using high-quality traces from DeepSeek-R1 as demonstrations.<n>We find that adding more exemplars consistently degrades accuracy, even when demonstrations are optimal.<n>We introduce Insight-to-solve (I2S), a sequential test-time procedure that turns demonstrations into explicit, reusable insights.
• Score: 58.02809208460186
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent reasoning LLMs (RLMs), especially those trained with verifier-based reinforcement learning, often perform worse with few-shot CoT than with direct answering. We revisit this paradox using high-quality reasoning traces from DeepSeek-R1 as demonstrations and find that adding more exemplars consistently degrades accuracy, even when demonstrations are optimal. A detailed analysis reveals two mechanisms behind this decline: (i) semantic misguidance, where high textual similarity leads the model to treat the target as the same as the exemplar and to copy intermediate steps verbatim; and (ii) strategy transfer failure, where the model struggles to extract useful reasoning strategies and apply them to target questions. Guided by these, we introduce Insight-to-Solve (I2S), a sequential test-time procedure that turns demonstrations into explicit, reusable insights and derives a target-specific reasoning trace; optionally, the reasoning is self-refined for coherence and correctness (I2S+). Extensive experiments on diverse benchmarks show that I2S and I2S+ consistently outperform both direct answering and test-time scaling baselines across open- and closed-source models. Even for GPT models, our method helps: on AIME'25, GPT-4.1 rises by +14.0%, and o1-mini improves by +2.7% on AIME and +1.7% on GPQA, indicating that in-context demonstrations can be harnessed effectively via insight-refine-solve framework.

Related papers

• CASHEW: Stabilizing Multimodal Reasoning via Iterative Trajectory Aggregation [6.356820150960838]
  We introduce two complementary approaches inspired by test-time scaling to stabilize vision-language models.<n>CASHEW is an inference-time framework that stabilizes reasoning by iteratively aggregating multiple candidate trajectories into higher-quality reasoning traces.<n>CASHEW-RL is trained using Group Sequence Policy Optimization (GSPO) with a composite reward that encourages correct answers grounded in minimal yet sufficient visual evidence.
  arXiv  Detail & Related papers  (2026-01-12T21:24:45Z)
• When Actions Teach You to Think: Reasoning-Action Synergy via Reinforcement Learning in Conversational Agents [2.689316553293938]
  Supervised fine-tuning (SFT) has emerged as one of the most effective ways to improve the performance of large language models (LLMs) in downstream tasks.<n>We propose a pipeline in which LLMs generate reasoning steps that guide both the invocation of tools and the final answer generation for conversational agents.
  arXiv  Detail & Related papers  (2025-12-12T04:44:40Z)
• GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning [53.894789613838654]
  We introduce SEED-Bench-R1, a benchmark with complex real-world videos requiring balanced perception and reasoning.<n>Using SEED-Bench-R1, we find that standard GRPO, while improving answer accuracy, often reduces logical coherence between reasoning steps and answers, with only a 57.9% consistency rate.<n>We propose GRPO-CARE, a consistency-aware RL framework optimizing both answer correctness and reasoning coherence without explicit supervision.
  arXiv  Detail & Related papers  (2025-06-19T08:49:13Z)
• When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs [19.354141845315276]
  Chain-of-thought reasoning can significantly degrade instruction-following accuracy.<n>This is the first work to systematically expose reasoning-induced failures in instruction-following.
  arXiv  Detail & Related papers  (2025-05-16T16:36:00Z)
• SEAL: Steerable Reasoning Calibration of Large Language Models for Free [58.931194824519935]
  Large Language Models (LLMs) have demonstrated compelling capabilities for complex reasoning tasks via the extended chain-of-thought (CoT) reasoning mechanism.<n>Recent studies reveal substantial redundancy in the CoT reasoning traces, which negatively impacts model performance.<n>We introduce SEAL, a training-free approach that seamlessly calibrates the CoT process, improving accuracy while demonstrating significant efficiency gains.
  arXiv  Detail & Related papers  (2025-04-07T02:42:07Z)
• R-PRM: Reasoning-Driven Process Reward Modeling [53.06844294668382]
  Process Reward Models (PRMs) have emerged as a promising solution by evaluating each reasoning step.<n>Existing PRMs typically output evaluation scores directly, limiting both learning efficiency and evaluation accuracy.<n>We propose Reasoning-Driven Process Reward Modeling (R-PRM)<n>R-PRM generates seed data from limited annotations, effectively bootstrapping our model's reasoning capabilities.
  arXiv  Detail & Related papers  (2025-03-27T09:23:08Z)
• Steamroller Problems: An Evaluation of LLM Reasoning Capability with Automated Theorem Prover Strategies [0.18416014644193066]
  We evaluate the performance of GPT4, GPT3.5 Turbo and Google's recent Gemini model on problems from a steamroller domain. We found that the models' performance when using the ATP reasoning strategies was comparable to one-shot chain of thought.
  arXiv  Detail & Related papers  (2024-07-17T22:49:23Z)
• MR-Ben: A Meta-Reasoning Benchmark for Evaluating System-2 Thinking in LLMs [55.20845457594977]
  Large language models (LLMs) have shown increasing capability in problem-solving and decision-making.<n>We present a process-based benchmark MR-Ben that demands a meta-reasoning skill.<n>Our meta-reasoning paradigm is especially suited for system-2 slow thinking.
  arXiv  Detail & Related papers  (2024-06-20T03:50:23Z)
• Chain of Evidences and Evidence to Generate: Prompting for Context Grounded and Retrieval Augmented Reasoning [3.117335706912261]
  Chain of Evidences (CoE) and Evidence to Generate (E2G) are built upon two unique strategies.<n>Instead of unverified reasoning claims, our innovative approaches leverage the power of "evidence for decision making"<n>Our framework consistently achieves remarkable results across various knowledge-intensive reasoning and generation tasks.
  arXiv  Detail & Related papers  (2024-01-11T09:49:15Z)
• Re-Reading Improves Reasoning in Large Language Models [87.46256176508376]
  We introduce a simple, yet general and effective prompting method, Re2, to enhance the reasoning capabilities of off-the-shelf Large Language Models (LLMs) Unlike most thought-eliciting prompting methods, such as Chain-of-Thought (CoT), Re2 shifts the focus to the input by processing questions twice, thereby enhancing the understanding process. We evaluate Re2 on extensive reasoning benchmarks across 14 datasets, spanning 112 experiments, to validate its effectiveness and generality.
  arXiv  Detail & Related papers  (2023-09-12T14:36:23Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.