On 6000 Qwen3-8B AIME traces, late-clustered moderate-to-severe backtracks are more common in incorrect outputs, enabling prefix-causal burst-aware filtering that outperforms fixed-length cutoffs at shallow and intermediate depths.
Seal: Steerable reasoning calibration of large language models for free.arXiv preprint arXiv:2504.07986, 2025a
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
VerifySteer selectively steers hidden states at paragraph boundaries using latent correctness signals to control verifier strictness and outperform baselines on ProcessBench and Hard2Verify with lower compute.
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
Answer tokens show forward drift and key-anchor focus when reading correct reasoning traces; a geometric-plus-semantic SRQ steering method boosts quantitative reasoning accuracy without training.
Dynamic Rollout Editing reduces overthinking in RL-trained LLMs by editing post-answer continuations in successful rollouts and preferring the edited versions within GRPO groups.
DyCon dynamically controls reasoning depth in LRMs by modeling evolving difficulty from step-level embeddings, reducing redundant steps across multiple benchmarks.
PreRL applies reward-driven updates to P(y) in pre-train space, uses Negative Sample Reinforcement to prune bad reasoning paths and boost reflection, and combines with standard RL in Dual Space RL to outperform baselines on reasoning tasks.
A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.