mlsanityData to 14 Oct 2026

Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces

arXiv 2605.29288 ↗First posted 28 May 2026cs.AIcited by 0

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Long chain-of-thought (CoT) traces are widely used as supervision for reasoning-oriented LLM SFT, yet answer-correct traces can still lead to markedly different fine-tuning outcomes. We study post-conclusion continuation in answer-correct long-CoT data: a continuation where the answer appears sufficiently supported, but the trace continues with additional reasoning that remains in the supervised target. To test its training effect, we use a delete-only editor to construct answer-preserving suffix removal and compare CoT-based SFT on the original and processed traces. We observe improved SFT outcomes after removing the editor-identified post-conclusion continuation, suggesting that this continuation is harmful to training in our setting. We therefore refer to this empirically supported phenomenon as harmful continuation. Beyond this intervention, we further characterize the removed post-conclusion continuation through uncertainty and hidden-state progress. We observe persistent local uncertainty together with weakened terminal-directional progress, forming an uncertainty--geometry mismatch. Finally, we instantiate Harmful Continuation Cut (HCC), a lightweight boundary proxy that approximates the editor-identified post-conclusion continuation boundary.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 14 of its 27 references that are in the map. First posted on arXiv 28 May 2026.

Builds on

14 of 27 references.

Closest ancestors · 3

Same direction, newest first

  1. 2026On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning2 Apr
  2. 2026Probing the Trajectories of Reasoning Traces in Large Language Models30 Jan
  3. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan

Other references · 11

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 297
  3. 2025The Best Instruction-Tuning Data are Those That Fitcited by 52
  4. 2025Deconstructing Long Chain-of-Thought: A Structured Reasoning Optimization Framework for Long CoT Distillationcited by 43
  5. 2025Not All Correct Answers Are Equal: Why Your Distillation Source Matterscited by 25
  6. 2026Which Reasoning Trajectories Teach Students to Reason Better? A Simple Metric of Informative Alignmentcited by 13
  7. 2025Shape of Thought: When Distribution Matters More than Correctness in Reasoning Taskscited by 11
  8. 2026Long-Chain Reasoning Distillation via Adaptive Prefix Alignmentcited by 6
  9. 2025Distilling the Essence: Efficient Reasoning Distillation via Sequence Truncationcited by 5
  10. 2026Learning from Partial Chain-of-Thought via Truncated-Reasoning Self-Distillationcited by 3
  11. 2025Empowering Lightweight MLLMs with Reasoning via Long CoT SFTcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
  2. 2026Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Models
  3. 2025Early Stopping Chain-of-thoughts in Large Language Models
  4. 2026Reasoning-Trace Collapse: Evaluating the Loss of Explicit Reasoning During Fine-Tuning
  5. 2026Facts Don't Speak Louder than Words: The Behavioral Essence of Long-CoT Distillation
  6. 2026</think> Doesn't Stop Reasoning: Analysis of Spurious CoT Termination
  7. 2025Through the Valley: Path to Effective Long CoT Training for Small Language Models
  8. 2026SCA: Segment-Wise CoT Compression with Answer Alignment