DyCon dynamically controls reasoning depth in LRMs by modeling evolving difficulty from step-level embeddings, reducing redundant steps across multiple benchmarks.
Wait, we don’t need to “wait”! removing think- ing tokens improves reasoning efficiency.arXiv preprint arXiv:2506.08343
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
representative citing papers
Distilling and retrieving reusable reasoning skills lets LLMs solve coding and math problems with fewer tokens and higher accuracy.
LLM mathematical reasoning forms ordered, step-specific trajectories in representation space that already exist in base models, diverge for correct vs. incorrect solutions at late stages, and support both correctness prediction (ROC-AUC 0.87) and trajectory-based steering.
Graph-based pruning of redundant reflections in LLM chain-of-thought reduces average reasoning tokens by 42% while preserving or improving accuracy.
On-policy reverse-KL self-distillation from a “be concise” teacher cuts reasoning length up to 56% while preserving or slightly improving math accuracy across several models.
A pipeline combining a question-aware formula-preserving compressor, mixed-ratio SFT, and hierarchical RL compresses CoT on math benchmarks to ~27% of original tokens while preserving or slightly improving accuracy.
Adding controlled noise and irrelevant persona contexts across training and testing stages for strong LLMs yields better reasoning and efficiency than high-quality data alone, reaching 76.7% on AIME24/25 with Qwen2.5-32B.
ACOER applies adaptive correct-only efficiency rewards in GRPO to avoid reward collapse, yielding higher accuracy and over 60% fewer tokens on math reasoning benchmarks.
Suppressing anthropomorphic reflection markers via prompt and token interventions preserves or improves LLM reasoning performance on four benchmarks while models continue marker-free verification.
DTSR enables large reasoning models to dynamically assess chain-of-thought sufficiency via reflection signals and a sufficiency check, reducing reasoning length by 28.9-34.9% with minimal performance loss on Qwen3 models.
citing papers explorer
-
DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling
DyCon dynamically controls reasoning depth in LRMs by modeling evolving difficulty from step-level embeddings, reducing redundant steps across multiple benchmarks.
-
Thinking with Reasoning Skills: Fewer Tokens, More Accuracy
Distilling and retrieving reusable reasoning skills lets LLMs solve coding and math problems with fewer tokens and higher accuracy.
-
LLM Reasoning as Trajectories: Step-Specific Representation Geometry and Correctness Signals
LLM mathematical reasoning forms ordered, step-specific trajectories in representation space that already exist in base models, diverge for correct vs. incorrect solutions at late stages, and support both correctness prediction (ROC-AUC 0.87) and trajectory-based steering.
-
Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs
Graph-based pruning of redundant reflections in LLM chain-of-thought reduces average reasoning tokens by 42% while preserving or improving accuracy.
-
CRISP: Compressed Reasoning via Iterative Self-Policy Distillation
On-policy reverse-KL self-distillation from a “be concise” teacher cuts reasoning length up to 56% while preserving or slightly improving math accuracy across several models.
-
Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression
A pipeline combining a question-aware formula-preserving compressor, mixed-ratio SFT, and hierarchical RL compresses CoT on math benchmarks to ~27% of original tokens while preserving or slightly improving accuracy.
-
Input-Time Scaling: Adding Noise and Irrelevance into Less-Is-More Drastically Improves Reasoning Performance and Efficiency
Adding controlled noise and irrelevant persona contexts across training and testing stages for strong LLMs yields better reasoning and efficiency than high-quality data alone, reaching 76.7% on AIME24/25 with Qwen2.5-32B.
-
Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards
ACOER applies adaptive correct-only efficiency rewards in GRPO to avoid reward collapse, yielding higher accuracy and over 60% fewer tokens on math reasoning benchmarks.
-
Revisiting Anthropomorphic Reflection Markers in Large Language Model Reasoning
Suppressing anthropomorphic reflection markers via prompt and token interventions preserves or improves LLM reasoning performance on four benchmarks while models continue marker-free verification.
-
When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning
DTSR enables large reasoning models to dynamically assess chain-of-thought sufficiency via reflection signals and a sufficiency check, reducing reasoning length by 28.9-34.9% with minimal performance loss on Qwen3 models.