---
title: "Confidence Leaps in LLM Reasoning: Early Stopping and"
description: "Building on this, they propose an early-stopping rule that needs no training and no access to model internals: halt once any… 17 references."
authors: ["Pavel Tikhonov", "Ivan Oseledets", "Elena Tutubalina"]
year: 2026
venue: "Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)"
source: "https://aclanthology.org/2026.eacl-short.43/"
doi: "10.18653/v1/2026.eacl-short.43"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/tikhonov-2026-confidence-leaps-llm-reasoning-early/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/tikhonov-2026-confidence-leaps-llm-reasoning-early/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/25/) / Confidence Leaps in LLM Reasoning: Early Stopping and Cross-Model Transfer

# Confidence Leaps in LLM Reasoning: Early Stopping and Cross-Model Transfer

**Authors:** [Pavel Tikhonov](https://md.research.elamur.ai/free/en/authors/pavel-tikhonov/), [Ivan Oseledets](https://md.research.elamur.ai/free/en/authors/ivan-oseledets/) and [Elena Tutubalina](https://md.research.elamur.ai/free/en/authors/elena-tutubalina/) · **Year:** 2026 · **Venue:** [Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)](https://md.research.elamur.ai/free/en/venues/eacl/) ([EACL 2026](https://md.research.elamur.ai/free/en/venues/eacl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.eacl-short.43/) · [PDF](https://aclanthology.org/2026.eacl-short.43.pdf) · [DOI 10.18653/v1/2026.eacl-short.43](https://doi.org/10.18653/v1/2026.eacl-short.43) · [OpenAlex](https://openalex.org/W7140127370)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary

Reasoning LLMs are often assumed to become more certain gradually as they think. This paper tests that assumption on multiple-choice tasks and finds that certainty often arrives abruptly. The authors split chain-of-thought traces from Qwen3-32B, QwQ-32B and DeepSeek-R1 Distill Llama-70B into chunks at shift markers such as "Wait" or "Hmm". After each chunk, they force an answer and record the probabilities over options A–D on GPQA and LogiQA. When a model switches to the right option, that option's probability usually jumps sharply. The authors call this a "confidence leap", and it can occur at almost any position in the trace. Building on this, they propose an early-stopping rule that needs no training and no access to model internals: halt once any option's probability rises by more than a threshold between consecutive chunks. Where a leap occurs, the rule cuts reasoning tokens by about a fifth to a third without losing accuracy. Reasoning up to the leap also helps a different model, gpt-oss-20b, which suggests leaps carry transferable meaning. Code and data are public.

## Key points

- On GPQA, Qwen3-32B's probability for the correct answer rises by +0.69 on average when it switches to that answer. The distribution of changes is bimodal: large positive jumps occur almost only for correct switches.
- At accuracy-optimal thresholds, token savings on traces where a leap occurs range from 19.2% to 32.7% (e.g., 31.5% for Qwen3-32B on GPQA with unchanged accuracy). The share of traces where the rule fires ranges from 3.5% to 54.0%.
- DeepSeek-R1 Distill Llama-70B on GPQA gains +8.7 accuracy points (68.9% → 77.6%) while saving 24.2% of tokens. This hints that late reasoning can sometimes hurt.
- In cross-model transfer from Qwen3-32B to gpt-oss-20b, direct-answer accuracy rises from 33.33% (two chunks before the leap) to 41.79% (one chunk before) to 70.59% (at the leap chunk). Agreement with the source model rises from 44.78% to 88.24%.
- Letting gpt-oss-20b continue for 128 tokens adds +14.93 points from the chunk before the leap but only +1.87 from the leap chunk itself. Limitations: the method requires a finite answer set and relies on trigger-word chunking.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Skolkovo Institute of Science and Technology](https://md.research.elamur.ai/free/en/universities/skolkovo-institute-of-science-and-technology/) — University, Skolkovo, Russia · 5 works · 142 citations
- [AIRI - Artificial Intelligence Research Institute](https://md.research.elamur.ai/free/en/ngos/airi-artificial-intelligence-research-institute/) — Nonprofit, Moscow, Russia · 5 works · 29 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [🧮 Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/) · 1,426 works

## Cite

Cite as: Pavel Tikhonov, Ivan Oseledets and Elena Tutubalina. 2026. Confidence Leaps in LLM Reasoning: Early Stopping and Cross-Model Transfer. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{tikhonov2026confidence,
  title = {{Confidence Leaps in LLM Reasoning: Early Stopping and Cross-Model Transfer}},
  author = {Pavel Tikhonov, Ivan Oseledets and Elena Tutubalina},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-short.43},
  url = {https://doi.org/10.18653/v1/2026.eacl-short.43}
}
```

- [← previous Conceptualising love in Amharic](https://md.research.elamur.ai/free/en/papers/kacprzak-2026-conceptualising-love-amharic/)
- [next → Confidence-Aware Ranker Ensembles for Robust In-Context Knowledge Editing](https://md.research.elamur.ai/free/en/papers/nair-2026-confidence-aware-ranker-ensembles-robust/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/25/)

## References

17 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anum Afzal, Florian Matthes, Gal Chechik, and Yftah Ziser. 2025. Knowing before saying: Llm representations encode information about chain-of-thought success before completion. arXiv preprint arXiv:2505.24362. [find](https://scholar.google.com/scholar?q=Anum+Afzal%2C+Florian+Matthes%2C+Gal+Chechik%2C+and+Yftah+Ziser%2E+2025%2E+Knowing+before+saying%3A+Llm+representations+encode+information+about+chain%2Dof%2Dthought+success+before+completion%2E+arXiv+preprint+arXiv%3A25)
2. Yifan Chen, Zhaojiang Liu, Shuohang Li, and 1 others. 2024. Rethinking the cost of reasoning: Adaptive depth in chain-of-thought generation. arXiv preprint arXiv:2401.02412. [find](https://scholar.google.com/scholar?q=Yifan+Chen%2C+Zhaojiang+Liu%2C+Shuohang+Li%2C+and+1+others%2E+2024%2E+Rethinking+the+cost+of+reasoning%3A+Adaptive+depth+in+chain%2Dof%2Dthought+generation%2E+arXiv+preprint+arXiv%3A2401%2E02412%2E)
3. Yao Fu, Xiaochuang Tang, Hao Peng, and 1 others. 2023. Chain-of-thought early exit: When to stop thinking in large language models. arXiv preprint arXiv:2310.04369. — [arXiv:2310.04369](https://arxiv.org/abs/2310.04369) [find](https://scholar.google.com/scholar?q=Yao+Fu%2C+Xiaochuang+Tang%2C+Hao+Peng%2C+and+1+others%2E+2023%2E+Chain%2Dof%2Dthought+early+exit%3A+When+to+stop+thinking+in+large+language+models%2E+arXiv+preprint+arXiv%3A2310%2E04369%2E+%E2%80%94+%5BarXiv%3A2310%2E04369%5D%28https%3A%2F%2Farxiv%2E)
4. Andrey Galichin, Alexey Dontsov, Polina Druzhinina, Anton Razzhigaev, Oleg Y. Rogov, Elena Tutubalina, and Ivan Oseledets. 2026. I have covered all the bases here: Interpreting reasoning features in large language models via sparse autoencoders. In Proceedings of the AAAI Conference on Artificial In… [find](https://scholar.google.com/scholar?q=Andrey+Galichin%2C+Alexey+Dontsov%2C+Polina+Druzhinina%2C+Anton+Razzhigaev%2C+Oleg+Y%2E+Rogov%2C+Elena+Tutubalina%2C+and+Ivan+Oseledets%2E+2026%2E+I+have+covered+all+the+bases+here%3A+Interpreting+reasoning+features+in+l)
5. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). arXiv preprint arXiv:2501.12948.
6. Andrew K. Lampinen, Ishita Dasgupta, James L. Mc-Clelland, and Noah D. Goodman. 2024. [Can language models learn from explanations in context?](https://md.research.elamur.ai/free/en/papers/lampinen-2022-can-language-models-learn-explanations/) arXiv preprint arXiv:2401.10968.
7. Tom Lanham, Yuhuai Wu, and Noah D. Goodman. 2023. [Measuring faithfulness in chain-of-thought reasoning](https://md.research.elamur.ai/free/en/papers/lanham-2023-measuring-faithfulness-chain-thought-reasoning/). arXiv preprint arXiv:2307.13702.
8. Jian Liu, Leyang Cui, Hanmeng Liu, Dandan Huang, Yile Wang, and Yue Zhang. 2020. [Logiqa: A challenge dataset for machine reading comprehension with logical reasoning](https://md.research.elamur.ai/free/en/papers/liu-2020-logiqa-challenge-dataset-machine-reading/). arXiv preprint arXiv:2007.08124.
9. Xin Liu and Lu Wang. 2025. [Answer convergence as a signal for early stopping in reasoning](https://md.research.elamur.ai/free/en/papers/liu-2025-answer-convergence-as-signal-early/). arXiv preprint arXiv:2506.02536.
10. Minjia Mao, Bowen Yin, Yu Zhu, and Xiao Fang. 2025. Early stopping chain-of-thoughts in large language models. arXiv preprint arXiv:2509.14004. — [arXiv:2509.14004](https://arxiv.org/abs/2509.14004) [find](https://scholar.google.com/scholar?q=Minjia+Mao%2C+Bowen+Yin%2C+Yu+Zhu%2C+and+Xiao+Fang%2E+2025%2E+Early+stopping+chain%2Dof%2Dthoughts+in+large+language+models%2E+arXiv+preprint+arXiv%3A2509%2E14004%2E+%E2%80%94+%5BarXiv%3A2509%2E14004%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2509%2E14004%29)
11. QwenTeam. 2025. Qwq-32b: Embracing the power of reinforcement learning. [https://qwen.ai/blog?id=](https://qwen.ai/blog?id=) 6aed6aa257238a0b6c77a6753f180350c2fecc4a& Accessed: from=research.research-list. 2025-10-04. [↗](https://qwen.ai/blog?id=)
12. David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. 2024. [Gpqa: A graduate-level google-proof q&a benchmark](https://md.research.elamur.ai/free/en/papers/rein-2023-gpqa-graduate-level-google-proof/). In First Conference on Language Modeling.
13. Charlie Snell, James Thorne, and Roger Grosse. 2023. Reducing the cost of reasoning via confidence-based early stopping. arXiv preprint arXiv:2310.08838. — [arXiv:2310.08838](https://arxiv.org/abs/2310.08838) [find](https://scholar.google.com/scholar?q=Charlie+Snell%2C+James+Thorne%2C+and+Roger+Grosse%2E+2023%2E+Reducing+the+cost+of+reasoning+via+confidence%2Dbased+early+stopping%2E+arXiv+preprint+arXiv%3A2310%2E08838%2E+%E2%80%94+%5BarXiv%3A2310%2E08838%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F231)
14. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). Advances in neural information processing systems, 35:24824– 24837.
15. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). arXiv preprint arXiv:2505.09388.
16. Anqi Zhang, Yulin Chen, Jane Pan, Chen Zhao, Aurojit Panda, Jinyang Li, and He He. 2025. Reasoning models know when they’re right: Probing hidden states for self-verification. arXiv preprint arXiv:2504.05419. [find](https://scholar.google.com/scholar?q=Anqi+Zhang%2C+Yulin+Chen%2C+Jane+Pan%2C+Chen+Zhao%2C+Aurojit+Panda%2C+Jinyang+Li%2C+and+He+He%2E+2025%2E+Reasoning+models+know+when+they%E2%80%99re+right%3A+Probing+hidden+states+for+self%2Dverification%2E+arXiv+preprint+arXiv%3A250)
17. Probability Change at the Critical Chunk A This section contains figures analyzing the change in probability for the correct answer at the identified critical chunk. *Figure 9: Distribution of probability change at the critical chunk for DeepSeek-R1 on LogiQA.* *Figure 5: Scatter plot of probability… [find](https://scholar.google.com/scholar?q=Probability+Change+at+the+Critical+Chunk+A+This+section+contains+figures+analyzing+the+change+in+probability+for+the+correct+answer+at+the+identified+critical+chunk%2E+%2AFigure+9%3A+Distribution+of+probabi)

*Updated 1 November 2026*