Challenge: Prior work on activation steering has focused on shaping reasoning traces, but it remains unclear how answer tokens actually read and integrate the reasoning to produce reliable outcomes.
Approach: They propose a training-free steering method that uses self-reading quality scores to guide inference toward benign self-readiness and away from uncertain and disorganized reading.
Outcome: The proposed method yields consistent accuracy gains in the reasoning traces generated by thinking LLMs.

Similar Papers

Tracing and Dissecting How LLMs Recall Factual Knowledge for Real World Questions (2025.acl-long)

Copied to clipboard

Challenge: Recent advances in large language models have shown promising ability to perform commonsense reasoning.
Approach: They propose a two-dimensional analysis framework that incorporates token back-tracing and token decoding to uncover how LLMs conduct factual knowledge recall.
Outcome: The proposed framework shows that LLMs lack relevant knowledge but struggle to select the most accurate information based on context during the retrieval and rerank phase.
Confidence Improves Self-Consistency in LLMs (2025.findings-acl)

Copied to clipboard

Challenge: Modern large language models (LLMs) demonstrate strong reasoning capabilities, driven in part by their capacity to generate a sequence of intermediate reasoning steps that lead them toward a final answer.
Approach: They propose a method that performs a weighted majority vote based on confidence scores obtained directly from the model.
Outcome: The proposed method outperforms self-consistency on nine models and four datasets, reducing the required number of reasoning paths by over 40% on average.
From Reasoning to Answer: Empirical, Attention-Based and Mechanistic Insights into Distilled DeepSeek R1 Models (2025.emnlp-main)

Copied to clipboard

Challenge: Large Reasoning Models generate explicit reasoning traces alongside final answers . the extent to which these traces influence answer generation remains unclear .
Approach: They conduct empirical evaluation of Large Reasoning Models that include explicit reasoning . they also show that answer tokens attend substantially to reasoning tokens .
Outcome: The results show that including explicit reasoning improves answer quality across domains . they also show that answer tokens attend substantially to reasoning tokens - the authors .
LLMs Faithfully and Iteratively Compute Answers During CoT: A Systematic Analysis With Multi-step Arithmetics (2026.findings-eacl)

Copied to clipboard

Challenge: Specifically, we examine when the LLMs’ answer is (pre)determined, especially before the CoT begins or after, and how strongly the information from CoT specifically has a causal effect on the final answer.
Approach: They examine when the LLMs’ answer is (pre)determined, especially before the CoT begins or after, and how strongly the information from CoT specifically has a causal effect on the final answer.
Outcome: The proposed model can generate reasoning chains while generating the reasoning chain on the fly.
Aligning What LLMs Do and Say: Towards Self-Consistent Explanations (2026.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) are often prompted to produce natural language explanations, but the features driving the answer are often different from those emphasized in their explanations.
Approach: They propose a large-scale benchmark linking model decisions with diverse explanations and attribution vectors across datasets, methods, and model families to address this gap.
Outcome: The proposed model generates an answer where the word NLP in the prompt has high feature importance.
Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Sampling (2025.naacl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) generate reasoning paths before answers, but lack a systematic approach to determine optimal number of samples or select the most faithful rationale.
Approach: They propose a framework that evaluates the quality of reasoning and consistency of answers for each generated sample and uses criteria-based stopping and weighted majority voting to guide early stopping decisions and rationale selection.
Outcome: The proposed framework outperforms existing methods while maintaining accuracy.
Interpretable Traces, Unexpected Outcomes: Investigating the Disconnect in Trace-Based Knowledge Distillation (2026.acl-long)

Copied to clipboard

Challenge: Recent advances in reasoning-oriented Large Language Models have been driven by the introduction of Chain-of-Thought (CoT) traces.
Approach: They propose to use CoT traces to guide model inference and serve as supervision signals for Knowledge Distillation to improve smaller models.
Outcome: The proposed model is based on a rule-based problem decomposition method and is valid for both semantic correctness and interpretability to the end user.
Self-Training Meets Consistency: Improving LLMs’ Reasoning with Consistency-Driven Rationale Evaluation (2025.naacl-long)

Copied to clipboard

Challenge: Existing approaches labeled rationales that produce correct answers as appropriate for training but one measure risks misjudging rationale quality, leading models to learn flawed reasoning patterns.
Approach: They propose a framework that evaluates rationales through follow-up questions and leverages this evaluation to guide its training.
Outcome: The proposed framework improves robustness and correctness of rationales and reasoning abilities compared to previous self-training approaches.
Neuro-Symbolic Integration Brings Causal and Reliable Reasoning Proofs (2025.findings-naacl)

Copied to clipboard

Challenge: a new framework for complex reasoning with LLMs is developed to improve reasoning proof accuracy and interpretability.
Approach: They propose to use LLMs to generate search logs that can be interpreted into human-readable reasoning proofs.
Outcome: The proposed framework improves reasoning accuracy but lacks interpretability due to black-box nature of the solvers.
“My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models (2024.findings-acl)

Copied to clipboard

Challenge: Multiple choice questions are one of the most popular evaluation formats for understanding the capabilities of autoregressive large language models (LLMs).
Approach: They evaluated how aligned first-token evaluation is with the text output along several dimensions, namely final option choice, refusal rate, choice distribution and robustness under prompt perturbation.
Outcome: The proposed evaluation methods are misaligned on all dimensions, reaching mismatch rates over 60%.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations