arxiv
PublishedMay 8, 2026 at 4:00 AM
—neutral
Measuring Evaluation-Context Divergence in Open-Weight LLMs: A Paired-Prompt Protocol with Pilot Evidence of Alignment-Pipeline-Specific Heterogeneity
Publisher summary· verbatim
arXiv:2605.06327v1 Announce Type: cross Abstract: Safety benchmarks are routinely treated as evidence about how a language model will behave once deployed, but this inference is fragile if behavior depends on whether a prompt looks like an evaluation. We define evaluation-context divergence as an ob
Models mentioned
02Related
05- arxivJul 29FusionML: Prefill, Not Decode - Mechanism and Boundaries of CPU+GPU Co-Execution on Unified-Memory Apple Silicon
- arxivJul 18Branching Policy Optimization: Sandbox-Native Language Agent Reinforcement Learning
- arxivJul 16A Shared Subcircuit Lets LLMs Count Down Across Tasks
- techcrunchJun 30Vibe-coding platform Base44 launches own model as AI startups seek defensibility
- arxivMay 28A Benchmark Construction and Evaluation Framework for Specialist Domains: Case Study on Defense-related Documents
Stay posted· Newsletter
A 5-min weekly brief — top movers, price watch, story of the week.
Discussion
No replies yet. Be first.
Related coverage
More from ARXIV
arxivFine-Tuning of Transformer models with Frames1darxivFeature Transformation Enhanced Jacobi Polynomial Graph Filtering for Graph Anomaly Detection1darxivNaive Prompt Optimization: Rethinking the Need for Complex Prompt Search1darxivSyntax vs. Semantics: How Transformers Learn Deep Dependencies1dThe Bubble Brief
WEEKLYRead safety insights every Tuesday — top movers, new releases, story of the week.
Originally published on arxiv ↗