Separable Pathways for Causal Reasoning: How Architectural Scaffolding Enables Hypothesis-Space Restructuring in LLM Agents

发布

2026年04月23日

采集 2026年04月23日 00:00

行业动态 6.5 分 — Addresses how LLM agents can restructure hypothesis space for causal discovery. Important for scientific agent capabilities.

原文： arXiv

评分 6.5 · 来源： · 发布于

评分依据：Addresses how LLM agents can restructure hypothesis space for causal discovery. Important for scientific agent capabilities.

From Signal Degradation to Computation Collapse: Uncovering the Two Failure Modes of LLM Quantization

Bootstrapping Post-training Signals for Open-ended Tasks via Rubric-based Self-play on Pre-training Text