# When Built-in Thinking Helps and Hurts: Constraint-Level Error Shifts in Instruction Following

arXiv: https://arxiv.org/abs/2606.09662 · HTML: https://mlsanity.com/paper/2606.09662/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Mathematical reasoning, one of 17 directions in LLM reasoning.
- Builds on 11 of its 11 references that are in the map.
- First posted on arXiv 8 Jun 2026.

## Placed in

- LLM reasoning / [Mathematical reasoning](https://mlsanity.com/lr/mathematical-reasoning/)

## Builds on (most cited first, 11 of 11)

- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [MS MARCO: A Human Generated MAchine Reading COmprehension Dataset](https://mlsanity.com/paper/1611.09268/) (2016)
- [Show Your Work: Scratchpads for Intermediate Computation with Language Models](https://mlsanity.com/paper/2112.00114/) (2021)
- [To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning](https://mlsanity.com/paper/2409.12183/) (2024)
- [Benchmarking Complex Instruction-Following with Multiple Constraints Composition](https://mlsanity.com/paper/2407.03978/) (2024)
- [InFoBench: Evaluating Instruction Following Ability in Large Language Models](https://mlsanity.com/paper/2401.03601/) (2024)
- [When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs](https://mlsanity.com/paper/2505.11423/) (2025)
- [Thinking LLMs: General Instruction Following with Thought Generation](https://mlsanity.com/paper/2410.10630/) (2024)
- [ReasonIF: Large Reasoning Models Fail to Follow Instructions During Reasoning](https://mlsanity.com/paper/2510.15211/) (2025)
- [Can LLMs Simulate Personas with Reversed Performance? A Systematic Investigation for Counterfactual Instruction Following in Math Reasoning Context](https://mlsanity.com/paper/2504.06460/) (2025)

## Abstract (arXiv)

Large reasoning models (LRMs) often improve math and coding performance, but their effect on instruction following is unclear. We study IFEval with Qwen3 models (1.7B-32B), using same-weights Thinking ON/OFF controls; four Hunyuan models provide directional cross-family support. Aggregate pass-rate changes are small (-0.55 to -3.52 pp), yet 10-20% of prompts switch between pass and fail across modes, suggesting that thinking changes the pattern of errors--some prompts improve while others worsen--rather than uniformly degrading performance. Under a post-hoc Qwen3-derived grouping, constraint types separate into Planning (global counting, structure, coordination), which improves at the class level under thinking, and Precision (exact local form), which consistently worsens; the class-level Planning/Precision sign pattern holds directionally for all four Hunyuan models despite Hunyuan's opposite aggregate direction. Thinking also changes final-answer length; matched-length analyses substantially reduce the Precision drop, but a residual penalty remains. Analyzing thinking traces with a cross-encoder relevance metric reveals three patterns: Neutral shows a positive relevance-compliance link (r approximately 0.15); Planning shows near-zero predictive correlation (r approximately 0.02) despite measurable trace engagement, consistent with an execution gap between CE-measured trace relevance and final-answer compliance; Precision shows a small negative correlation (r approximately -0.05), with failing instances having higher mean relevance than passing ones. Activation patching across four model sizes (1.7B-14B) shows that Precision flip instances are more often restored than Planning flip instances (32-58% vs. 14-40% mean layer-restoration), with the largest gap at 14B (about 30 pp).