# Large Language Models Decide Early and Explain Later

arXiv: https://arxiv.org/abs/2604.22266 · HTML: https://mlsanity.com/paper/2604.22266/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 25 of its 26 references that are in the map.
- First posted on arXiv 24 Apr 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 25 of 25)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Least-to-Most Prompting Enables Complex Reasoning in Large Language Models](https://mlsanity.com/paper/2205.10625/) (2022)
- [Solving Quantitative Reasoning Problems with Language Models](https://mlsanity.com/paper/2206.14858/) (2022)
- [Measuring Faithfulness in Chain-of-Thought Reasoning](https://mlsanity.com/paper/2307.13702/) (2023)
- [Attention is not not Explanation](https://mlsanity.com/paper/1908.04626/) (2019)
- [Humanity's Last Exam](https://mlsanity.com/paper/2501.14249/) (2025)
- [MathArena: Evaluating LLMs on Uncontaminated Math Competitions](https://mlsanity.com/paper/2505.23281/) (2025)
- [Can language models learn from explanations in context?](https://mlsanity.com/paper/2204.02329/) (2022)
- [MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning](https://mlsanity.com/paper/2205.00445/) (2022)
- [Multi-Step Reasoning with Large Language Models, a Survey](https://mlsanity.com/paper/2407.11511/) (2024)
- [Thought Anchors: Which LLM Reasoning Steps Matter?](https://mlsanity.com/paper/2506.19143/) (2025)
- [What Evidence Do Language Models Find Convincing?](https://mlsanity.com/paper/2402.11782/) (2024)
- [Reasoning Models Better Express Their Confidence](https://mlsanity.com/paper/2505.14489/) (2025)
- [Internal Consistency and Self-Feedback in Large Language Models: A Survey](https://mlsanity.com/paper/2407.14507/) (2024)
- [Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?](https://mlsanity.com/paper/2506.18183/) (2025)
- [RepLiQA: A Question-Answering Dataset for Benchmarking LLMs on Unseen Reference Content](https://mlsanity.com/paper/2406.11811/) (2024)
- [CoT-UQ: Improving Response-wise Uncertainty Quantification in LLMs with Chain-of-Thought](https://mlsanity.com/paper/2502.17214/) (2025)
- [Recurrent Confidence Chain: Temporal-Aware Uncertainty Quantification in Large Language Models](https://mlsanity.com/paper/2601.13368/) (2026)
- [Confidence over Time: Confidence Calibration with Temporal Logic for Large Language Model Reasoning](https://mlsanity.com/paper/2601.13387/) (2026)

## Built on by (most cited first, 2 of 2)

- [Shorter Reasoning, Earlier Answers? An Evaluation of Reasoning Interfaces](https://mlsanity.com/paper/2608.03401/) (2026)
- [LLMs Should Not Yet Be Credited with Decision Explanation](https://mlsanity.com/paper/2605.01164/) (2026)

## Abstract (arXiv)

Large Language Models often achieve strong performance by generating long intermediate chain-of-thought reasoning. However, it remains unclear when a model's final answer is actually determined during generation. If the answer is already fixed at an intermediate stage, subsequent reasoning tokens may constitute post-decision explanation, increasing inference cost and latency without improving correctness. We study the evolution of predicted answers over reasoning steps using forced answer completion, which elicits the model's intermediate predictions at partial reasoning prefixes. Focusing on Qwen3-4B and averaging results across all datasets considered, we find that predicted answers change in only 32% of queries. Moreover, once the final answer switch occurs, the model generates an average of 760 additional reasoning tokens per query, accounting for a substantial fraction of the total reasoning budget. Motivated by these findings, we investigate early stopping strategies that halt generation once the answer has stabilized. We show that simple heuristics, including probe-based stopping, can reduce reasoning token usage by 500 tokens per query while incurring only a 2% drop in accuracy. Together, our results indicate that a large portion of chain-of-thought generation is redundant and can be reduced with minimal impact on performance.