# From Hallucination to Structure Snowballing: The Alignment Tax of Constrained Decoding in LLM Reflection

arXiv: https://arxiv.org/abs/2604.06066 · HTML: https://mlsanity.com/paper/2604.06066/

## Key facts

- Cited by 1 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 15 of its 15 references that are in the map.
- First posted on arXiv 7 Apr 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)
- LLM alignment / RLHF / preference optimisation / [LLM preference optimization methods](https://mlsanity.com/alignment/llm-preference-optimization-methods/)

## Builds on (most cited first, 15 of 15)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Self-Refine: Iterative Refinement with Self-Feedback](https://mlsanity.com/paper/2303.17651/) (2023)
- [Tree of Thoughts: Deliberate Problem Solving with Large Language Models](https://mlsanity.com/paper/2305.10601/) (2023)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [How Language Model Hallucinations Can Snowball](https://mlsanity.com/paper/2305.13534/) (2023)
- [REFINER: Reasoning Feedback on Intermediate Representations](https://mlsanity.com/paper/2304.01904/) (2023)
- [Efficient Guided Generation for Large Language Models](https://mlsanity.com/paper/2307.09702/) (2023)
- [ROSCOE: A Suite of Metrics for Scoring Step-by-Step Reasoning](https://mlsanity.com/paper/2212.07919/) (2022)
- [Compositional Questions Do Not Necessitate Multi-hop Reasoning](https://mlsanity.com/paper/1906.02900/) (2019)
- [LLMs cannot find reasoning errors, but can correct them given the error location](https://mlsanity.com/paper/2311.08516/) (2023)
- [MAF: Multi-Aspect Feedback for Improving Reasoning in Large Language Models](https://mlsanity.com/paper/2310.12426/) (2023)
- [Decomposing LLM Self-Correction: The Accuracy-Correction Paradox and Error Depth Hypothesis](https://mlsanity.com/paper/2601.00828/) (2025)

## Built on by (most cited first, 1 of 1)

- [Repair, Not Improvement: Decomposing Constrained Decoding in Tool-Call Abstention](https://mlsanity.com/paper/2608.13959/) (2026)

## Abstract (arXiv)

Intrinsic self-correction in Large Language Models (LLMs) frequently fails in open-ended reasoning tasks due to ``hallucination snowballing,'' a phenomenon in which models recursively justify early errors during free-text reflection. While structured feedback can mitigate this issue, existing approaches often rely on externally trained critics or symbolic tools, reducing agent autonomy. This study investigates whether enforcing structured reflection purely through Outlines-based constrained decoding can disrupt error propagation without additional training. Evaluating an 8-billion-parameter model (Qwen3-8B), we show that simply imposing structural constraints does not improve self-correction performance. Instead, it triggers a new failure mode termed ``structure snowballing.'' We find that the cognitive load required to satisfy strict formatting rules pushes the model into formatting traps. This observation helps explain why the agent achieves near-perfect superficial syntactic alignment yet fails to detect or resolve deeper semantic errors. These findings expose an ``alignment tax'' inherent to constrained decoding, highlighting a tension between structural granularity and internal model capacity in autonomous workflows. Code and raw logs are available in the GitHub repository: https://github.com/hongxuzhou/agentic_llm_structured_self_critique.