Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models

• URL: http://arxiv.org/abs/2601.14290v1
• Date: Wed, 14 Jan 2026 14:39:08 GMT
• Title: Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models
• Authors: Aradhya Dixit, Tianxi Liang, Jai Telang,
• Abstract summary: Small Language Models (SLMs, under 10B parameters) are attractive for private, on-device deployment.<n>We introduce Verifier-Guided Distillation, a training protocol that transfers the process of error repair.<n>We show that latent verification behavior can emerge in small models, enabling them to occasionally stop, detect contradictions, and revise earlier assumptions.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Small Language Models (SLMs, under 10B parameters) are attractive for private, on-device deployment, yet they frequently fail on strict constraint-satisfaction problems due to linear, overconfident reasoning traces that do not recover from early mistakes. We introduce Verifier-Guided Distillation, a training protocol that transfers the process of error repair - explicit conflict detection and backtracking - rather than only correct final answers. By training a 7B model on verified reasoning traces that include mistakes and self-corrections, we show that latent verification behavior can emerge in small models, enabling them to occasionally stop, detect contradictions, and revise earlier assumptions.

Related papers

• Online Learnability of Chain-of-Thought Verifiers: Soundness and Completeness Trade-offs [34.168578803480116]
  We propose an online learning framework for learning chain-of-thought verifiers.<n>We show how our learned verifiers can be used to boost the accuracy of a collection of weak provers.
  arXiv  Detail & Related papers  (2026-03-03T21:50:14Z)
• InT: Self-Proposed Interventions Enable Credit Assignment in LLM Reasoning [32.274434679047395]
  Outcome-reward reinforcement learning (RL) has proven effective at improving the reasoning capabilities of large language models (LLMs)<n>Standard RL assigns credit only at the level of the final answer, penalizing entire reasoning traces when the outcome is incorrect.<n>We introduce Intervention Training (InT), a training paradigm in which the model performs fine-grained credit assignment on its own reasoning traces.
  arXiv  Detail & Related papers  (2026-01-20T18:15:38Z)
• From Denoising to Refining: A Corrective Framework for Vision-Language Diffusion Model [72.73512218682187]
  We introduce ReDiff, a refining-enhanced diffusion framework that teaches the model to identify and correct its own errors.<n>Our approach features a two-stage training process: first, we instill a foundational revision capability by training the model to revise synthetic errors; second, we implement a novel online self-correction loop.<n>This mistake-driven learning endows the model with the crucial ability to revisit and refine its already generated output, effectively breaking the error cascade.
  arXiv  Detail & Related papers  (2025-10-22T06:58:55Z)
• AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models [62.70575022567081]
  We propose AdvChain, an alignment paradigm that teaches models dynamic self-correction through adversarial CoT tuning.<n>Our work establishes a new direction for building more robust and reliable reasoning models.
  arXiv  Detail & Related papers  (2025-09-29T04:27:23Z)
• Probing for Arithmetic Errors in Language Models [86.8227317662622]
  Internal activations in language models can be used to detect arithmetic errors.<n>We show that simple probes can accurately decode both the model's predicted output and the correct answer from hidden states.<n>We train lightweight error detectors that predict model correctness with over 90% accuracy.
  arXiv  Detail & Related papers  (2025-07-16T16:27:50Z)
• Language Models can perform Single-Utterance Self-Correction of Perturbed Reasoning [4.768151813962547]
  Large Language Models (LLMs) have demonstrated impressive mathematical reasoning capabilities.<n>Their performance remains brittle to minor variations in problem description and prompting strategy.<n>To better understand self-correction capabilities of recent models, we conduct experiments measuring models' ability to self-correct synthetics.
  arXiv  Detail & Related papers  (2025-06-18T21:35:44Z)
• Technical Report for Egocentric Mistake Detection for the HoloAssist Challenge [5.257305312436567]
  We introduce an online mistake detection framework that handles both procedural and execution errors.<n>Upon detecting an error, we use a large language model (LLM) to generate explanatory feedback.<n>Experiments on the HoloAssist benchmark confirm the effectiveness of our approach.
  arXiv  Detail & Related papers  (2025-06-06T15:39:09Z)
• Subtle Errors in Reasoning: Preference Learning via Error-injected Self-editing [59.405145971637204]
  We propose a novel preference learning framework called eRror-Injected Self-Editing (RISE)<n>RISE injects predefined subtle errors into pivotal tokens in reasoning or steps to construct hard pairs for error mitigation.<n>Experiments validate the effectiveness of RISE, with preference learning on Qwen2-7B-Instruct yielding notable improvements of 3.0% on GSM8K and 7.9% on MATH with only 4.5K training samples.
  arXiv  Detail & Related papers  (2024-10-09T07:43:38Z)
• SINDER: Repairing the Singular Defects of DINOv2 [61.98878352956125]
  Vision Transformer models trained on large-scale datasets often exhibit artifacts in the patch token they extract. We propose a novel fine-tuning smooth regularization that rectifies structural deficiencies using only a small dataset.
  arXiv  Detail & Related papers  (2024-07-23T20:34:23Z)
• PREGO: online mistake detection in PRocedural EGOcentric videos [49.72812518471056]
  We propose PREGO, the first online one-class classification model for mistake detection in egocentric videos. PREGO is based on an online action recognition component to model the current action, and a symbolic reasoning module to predict the next actions. We evaluate PREGO on two procedural egocentric video datasets, Assembly101 and Epic-tent, which we adapt for online benchmarking of procedural mistake detection.
  arXiv  Detail & Related papers  (2024-04-02T13:27:28Z)
• Understanding and Mitigating Classification Errors Through Interpretable Token Patterns [58.91023283103762]
  Characterizing errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors. We propose to discover those patterns of tokens that distinguish correct and erroneous predictions. We show that our method, Premise, performs well in practice.
  arXiv  Detail & Related papers  (2023-11-18T00:24:26Z)

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.