ProBackend
Model Performance & Safety

Model Performance & Safety

Articles on AI model performance shifts, safety guardrails, fallback behavior, and unintended restrictions after updates.

model performance safety2 days ago5 min

Hugging Face Brings Private ASR Benchmarks Online To Reduce Benchmaxxing

Appen Inc. and DataoceanAI donate private English speech datasets to the Open ASR Leaderboard. A new toggle and Rank Δ show how model rankings shift when private data is included, while the default Average WER stays on public sets.

model performance safetyAug 7, 20263 min

GLM-5.2 and the Open-Weight Safety Black Hole

A new SaferAI report finds Z.ai's open-weight GLM-5.2 approaches frontier AI capabilities while lacking key safety mitigations, renewing concerns that powerful open models could outpace governance and regulatory oversight.