On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning

• URL: http://arxiv.org/abs/2605.05438v1
• Date: Wed, 06 May 2026 20:54:30 GMT
• Title: On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning
• Abstract summary: We demonstrate that fine-tuning Gemma 270M on transitivity and d-separation tasks without semantic loss results in 100% collapse rate.<n>We propose a semantic loss function with graph-based logical constraints and dynamic scheduling that prevents this collapse.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Standard fine-tuning of transformer models on causal reasoning tasks leads to catastrophic model collapse, where models learn trivial solutions such as always predicting "Yes" or "No" regardless of input structure. We demonstrate that fine-tuning Gemma 270M on transitivity and d-separation tasks without semantic loss results in 100% collapse rate, with models achieving misleadingly high accuracy (73.9%) while learning no causal reasoning. We propose a semantic loss function with graph-based logical constraints and dynamic lambda scheduling that prevents this collapse. Our approach achieves 70.4% accuracy on transitivity tasks and 68.6% on d-separation tasks with stable, context-dependent predictions, representing a 42.7% improvement over collapsed baselines. Adversarial evaluation on 1,000 structural reasoning samples shows semantic models achieve 67-70% accuracy while collapsed models fail catastrophically at 43-71%. We validate our findings through comprehensive benchmarking on 200,000+ evaluation samples across five model variants, demonstrating that semantic loss is essential and not optional, for stable causal reasoning in transformers.

Related papers

• State of Thought Enables Endogenous Reasoning [48.902117018115256]
  Test-time reasoning has emerged as a major approach to improving the capabilities of Large Language Models (LLMs)<n>We propose State of Thought, a new reasoning paradigm that enables endogenous reasoning in LLMs.<n>We show that SoT consistently improves mean-baseline accuracy while reducing generated tokens by 62.6% and end-to-end latency by 44.6%.
  arXiv  Detail & Related papers  (2026-09-13T05:55:03Z)
• Reason, Reward, Refine: Step-Level Errors Corrections with Structured Feedback for Physics Reasoning in Small Language Models [32.7643564019866]
  We propose a step-level reward framework that identifies the first reasoning error, generates targeted structured feedback, and trains the model to revise its solution via policy gradient with KL regularization.<n>Across five physics benchmarks, our framework delivers accuracy gains of 17-20% over CoT prompting and 10-16% over the strongest baseline, reduces calculation errors from 56.9% to 23.5%, and reduces miscomprehension errors from 22.3% to 12.0% in the best observed cases.
  arXiv  Detail & Related papers  (2026-07-06T15:16:10Z)
• Automated Proving of Shannon-Type Entropy Inequalities via Fine-Tuned Language Models and Guided Tree Search [50.16356451328644]
  Proving Shannon-type entropy inequalities is a fundamental task in information theory.<n>We investigate whether small-scale large language models, fine-tuned on atomic proof steps, can automate this process.<n>We show that GPT-5.5 solves 1.7% samples under zero-shot prompting while Psitip solves 33.3% samples.
  arXiv  Detail & Related papers  (2026-06-04T05:43:12Z)
• Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models [94.68358825189738]
  We propose a post-training framework for verifiable domains that jointly optimize prediction accuracy and reasoning quality.<n>We evaluate VPS on chess, a controlled testbed where reasoning steps can be deterministically verified against engine signals.<n>VPS preserves accuracy while significantly improving reasoning quality, reducing win-rate error by up to 30% and restoring consistency to near saturation.
  arXiv  Detail & Related papers  (2026-04-03T15:19:46Z)
• The Model Says Walk: How Surface Heuristics Override Implicit Constraints in LLM Reasoning [9.898274894485107]
  Large language models systematically fail when a salient surface cue conflicts with an unstated feasibility constraint.<n>We study this through a diagnose-measure-bridge-treat framework.
  arXiv  Detail & Related papers  (2026-03-30T21:36:09Z)
• Robust Reasoning Benchmark [2.622879924644665]
  We propose a perturbation pipeline consisting of 14 techniques to evaluate robustness of LLM reasoning.<n>We apply this pipeline to AIME 2024 dataset and evalute 8 state-of-the-art models on the resulting benchmark.<n>Our results indicate that open weight models ranging from 7B to 120B parameters and Claude Opus 4.6 exhibit accuracy decay on subsequent problems.
  arXiv  Detail & Related papers  (2026-03-26T22:19:33Z)
• Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference [0.0]
  Transformer-based language models are widely deployed for reasoning, yet their behavior under inference-timeity remains underexplored.<n>This work analyzes dropout-induced variability across 19 transformer models using MC Dropout with 100 forward passes per sample.
  arXiv  Detail & Related papers  (2026-03-18T15:04:26Z)
• When Shallow Wins: Silent Failures and the Depth-Accuracy Paradox in Latent Reasoning [16.505918019260964]
  We demonstrate that state-of-the-art models (Qwen2.5-Math-7B) achieve 61% accuracy through a mixture of reliable and unreliable predictions.<n>We show that 18.4% of correct predictions employ stable, faithful reasoning while 81.6% emerge through computationally inconsistent pathways.
  arXiv  Detail & Related papers  (2026-03-03T19:43:36Z)
• Brittlebench: Quantifying LLM robustness via prompt sensitivity [44.950999933205985]
  We introduce a theoretical framework for quantifying model sensitivity to prompt variants, or brittleness.<n>We design a novel evaluation pipeline, Brittlebench, to holistically evaluate the sensitivity of frontier models.<n>We find that semantics-preserving input perturbations can account for up to half of the performance variance for a given model.
  arXiv  Detail & Related papers  (2026-02-27T21:12:13Z)
• Km-scale dynamical downscaling through conformalized latent diffusion models [45.94979929172337]
  Dynamical downscaling is crucial for deriving high-resolution meteorological fields from coarse-scale simulations.<n>Generative Diffusion models (DMs) have recently emerged as powerful data-driven tools for this task.<n>However, DMs lack finite-sample guarantees against overconfident predictions, resulting in miscalibrated grid-point-level uncertainty estimates.<n>We tackle this issue by augmenting the downscaling pipeline with a conformal prediction framework.
  arXiv  Detail & Related papers  (2025-10-15T08:41:36Z)
• OmniEAR: Benchmarking Agent Reasoning in Embodied Tasks [52.87238755666243]
  We present OmniEAR, a framework for evaluating how language models reason about physical interactions, tool usage, and multi-agent coordination in embodied tasks.<n>We model continuous physical properties and complex spatial relationships across 1,500 scenarios spanning household and industrial domains.<n>Our systematic evaluation reveals severe performance degradation when models must reason from constraints.
  arXiv  Detail & Related papers  (2025-08-07T17:54:15Z)
• Benchmarking Reasoning Robustness in Large Language Models [76.79744000300363]
  We find significant performance degradation on novel or incomplete data.<n>These findings highlight the reliance on recall over rigorous logical inference.<n>This paper introduces a novel benchmark, termed as Math-RoB, that exploits hallucinations triggered by missing information to expose reasoning gaps.
  arXiv  Detail & Related papers  (2025-03-06T15:36:06Z)
• Conservative Prediction via Data-Driven Confidence Minimization [70.93946578046003]
  In safety-critical applications of machine learning, it is often desirable for a model to be conservative. We propose the Data-Driven Confidence Minimization framework, which minimizes confidence on an uncertainty dataset.
  arXiv  Detail & Related papers  (2023-06-08T07:05:36Z)
• Robustness and Accuracy Could Be Reconcilable by (Proper) Definition [109.62614226793833]
  The trade-off between robustness and accuracy has been widely studied in the adversarial literature. We find that it may stem from the improperly defined robust error, which imposes an inductive bias of local invariance. By definition, SCORE facilitates the reconciliation between robustness and accuracy, while still handling the worst-case uncertainty.
  arXiv  Detail & Related papers  (2022-02-21T10:36:09Z)

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.