Answering the Wrong Question: Reasoning Trace Inversion for Abstention in LLMs

• URL: http://arxiv.org/abs/2604.02230v1
• Date: Thu, 02 Apr 2026 16:23:46 GMT
• Title: Answering the Wrong Question: Reasoning Trace Inversion for Abstention in LLMs
• Abstract summary: Reasoning models have been shown to have worse abstention abilities.<n>Hallucinations resulting in failed abstention can be reinterpreted as LLMs answering the wrong question.<n>We develop a new class of state-of-the-art abstention methods called Trace Inversion.
• Score: 24.31388053995892
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: For Large Language Models (LLMs) to be reliably deployed, models must effectively know when not to answer: abstain. Reasoning models, in particular, have gained attention for impressive performance on complex tasks. However, reasoning models have been shown to have worse abstention abilities. Taking the vulnerabilities of reasoning models into account, we propose our Query Misalignment Framework. Hallucinations resulting in failed abstention can be reinterpreted as LLMs answering the wrong question (rather than answering a question incorrectly). Based on this framework, we develop a new class of state-of-the-art abstention methods called Trace Inversion. First, we generate the reasoning trace of a model. Based on only the trace, we then reconstruct the most likely query that the model responded to. Finally, we compare the initial query with the reconstructed query. Low similarity score between the initial query and reconstructed query suggests that the model likely answered the question incorrectly and is flagged to abstain. Extensive experiments demonstrate that Trace Inversion effectively boosts abstention performance in four frontier LLMs across nine abstention QA datasets, beating competitive baselines in 33 out of 36 settings.

Related papers

• PassiveQA: A Three-Action Framework for Epistemically Calibrated Question Answering via Supervised Finetuning [0.0]
  We study decision-aware query resolution under incomplete information, where a model must determine whether to Answer, Ask for clarification, or Abstain.<n>We propose PassiveQA, a three-action framework that aligns model behaviour with information sufficiency through supervised finetuning.<n>Our approach integrates structured information-state representations, knowledge graph-grounded context, and a finetuned planner that explicitly models missing variables and decision reasoning.
  arXiv  Detail & Related papers  (2026-04-06T09:54:16Z)
• Probing the Trajectories of Reasoning Traces in Large Language Models [4.599673637363014]
  We propose a protocol to probe the trajectories of reasoning traces in large language models.<n>We find that accuracy and decision commitment consistently increase as the percentage of provided reasoning tokens grows.<n>We show that trajectory probing provides diagnostics for efficient and safer deployment of reasoning models.
  arXiv  Detail & Related papers  (2026-01-30T16:45:16Z)
• MathSE: Improving Multimodal Mathematical Reasoning via Self-Evolving Iterative Reflection and Reward-Guided Fine-Tuning [20.82742383613536]
  Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in vision-language answering tasks.<n>Previous works have focused on fine-tuning on specialized mathematical datasets.<n>Method iteratively refines the model through cycles of inference, reflection, and reward-based feedback.<n>Results on MathVL-test surpass the leading open-source multimodal mathematical reasoning model QVQ.
  arXiv  Detail & Related papers  (2025-11-10T07:46:19Z)
• First Try Matters: Revisiting the Role of Reflection in Reasoning Models [66.39546876232512]
  We focus on reflective behaviours where the model has already produced an answer but continues reflecting before finalizing its output.<n>Our analysis reveals that reflections are predominantly confirmatory and rarely alter the model's initial answer.<n>We propose a question-aware early-stopping method that enhances inference-time token efficiency by stopping the reasoning process once a few plausible candidate answers are generated.
  arXiv  Detail & Related papers  (2025-10-09T14:57:10Z)
• ASTRO: Teaching Language Models to Reason by Reflecting and Backtracking In-Context [66.15505423059234]
  We introduce ASTRO, a framework for training language models to reason like search algorithms.<n>We apply ASTRO to the Llama 3 family of models and achieve absolute performance gains of 16.4% on MATH-500, 26.9% on AMC 2023, and 20.0% on AIME 2024.
  arXiv  Detail & Related papers  (2025-07-01T04:10:15Z)
• Maximally-Informative Retrieval for State Space Model Generation [59.954191072042526]
  We introduce Retrieval In-Context Optimization (RICO) to minimize model uncertainty for a particular query at test-time.<n>Unlike traditional retrieval-augmented generation (RAG), which relies on externals for document retrieval, our approach leverages direct feedback from the model.<n>We show that standard top-$k$ retrieval with model gradients can approximate our optimization procedure, and provide connections to the leave-one-out loss.
  arXiv  Detail & Related papers  (2025-06-13T18:08:54Z)
• AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions [32.871820908561936]
  AbstentionBench is a benchmark for holistically evaluating abstention across 20 diverse datasets.<n>We find that reasoning fine-tuning degrades abstention even for math and science domains.
  arXiv  Detail & Related papers  (2025-06-10T17:57:30Z)
• S$^2$R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning [51.84977135926156]
  We introduce S$2$R, an efficient framework that enhances LLM reasoning by teaching models to self-verify and self-correct during inference.<n>Our results demonstrate that Qwen2.5-math-7B achieves an accuracy improvement from 51.0% to 81.6%, outperforming models trained on an equivalent amount of long-CoT distilled data.
  arXiv  Detail & Related papers  (2025-02-18T13:40:22Z)
• SPaR: Self-Play with Tree-Search Refinement to Improve Instruction-Following in Large Language Models [88.29990536278167]
  We introduce SPaR, a self-play framework integrating tree-search self-refinement to yield valid and comparable preference pairs.<n>Our experiments show that a LLaMA3-8B model, trained over three iterations guided by SPaR, surpasses GPT-4-Turbo on the IFEval benchmark without losing general capabilities.
  arXiv  Detail & Related papers  (2024-12-16T09:47:43Z)
• Sufficient Context: A New Lens on Retrieval Augmented Generation Systems [19.238772793096473]
  Augmenting LLMs with context leads to improved performance across many applications.<n>We develop a new notion of sufficient context, along with a method to classify instances that have enough information to answer the query.<n>By stratifying errors based on context sufficiency, we find that larger models with higher baseline performance excel at answering queries when the context is sufficient.
  arXiv  Detail & Related papers  (2024-11-09T02:13:14Z)
• LLMs can learn self-restraint through iterative self-reflection [57.26854891567574]
  Large Language Models (LLMs) must be capable of dynamically adapting their behavior based on their level of knowledge and uncertainty associated with specific topics. This adaptive behavior, which we refer to as self-restraint, is non-trivial to teach. We devise a utility function that can encourage the model to produce responses only when it is confident in them.
  arXiv  Detail & Related papers  (2024-05-15T13:35:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.