Diagnosing and Mitigating Thinking Collapse in On-Policy Self-Distillation
On-Policy Self-Distillation (OPSD) has emerged as a crucial paradigm for enhancing and aligning Large Language Models (LLMs). However, in complex reasoning tasks, OPSD paradoxically degrades downstream performance. In this paper, we systematically investigate this pathology and identify a severe optimization trap we define as \textbf{Thinking Collapse} -- a sharp decline in the model's native intermediate reasoning behavior, measured by epistemic-token density (ET per 1k). Through entropy-based
Record details
Published: 12 July 2026
Source: arXiv cs.LG
Category: Research
Topics: Regulation · Healthcare · Finance, VC & PE
Retrieved: 14 July 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
From Siloed Algorithms to Compliance-First Agentic Platforms: A Multi-Layered Architecture for Hospital AI Systems
arXiv · 6 August 2026
Physiological and Semantic Patterns in Medical Teams Using an Intelligent Tutoring System
arXiv · 31 March 2026
Burden of disease scenarios for 204 countries and territories, 2022–2050: a forecasting analysis for the Global Burden of Disease Study 2021
OpenAlex · 1 May 2024
A governance model for the application of AI in health care
OpenAlex · 10 October 2019
FDA-ARGOS is a database with public quality-controlled reference genomes for diagnostic use and regulatory science
OpenAlex · 25 July 2019
Physicians’ Perceptions of Chatbots in Health Care: Cross-Sectional Web-Based Survey
OpenAlex · 9 February 2019
How to cite this record
ethics.ai (12 July 2026), “Diagnosing and Mitigating Thinking Collapse in On-Policy Self-Distillation,” evidence record 10211, https://ethics.ai/record/10211 (originally published by arXiv cs.LG).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.