Astra's big reasoning jump without chain of thought raises covert-reasoning concerns
JeffLadish · x · 2026-09-11
Safety researcher Jeff Ladish flags a concerning jump in Astra's reasoning ability with no chain of thought. Neel Nanda replicated the UK AI Security Institute's finding using his own private benchmark. Ladish argues this greatly increases how much covert reasoning the model could be doing — undermining chain-of-thought-based model monitoring.
More from Models
- Nanbeige4.3-33B benchmarked: 59 tok/s single stream, ~1,060 tok/s across 64 concurrent runs — MaziyarPanahi · 2026-10-26
- Only 2 of 25 full ChatGPT rewrites flagged 100% AI by detectors — paulnovosad · 2026-10-26
- Qwen 4 Reportedly Spotted in the Arena, Hints at Alibaba's Next-Gen Model Testing — AIFlow_ML · 2026-10-26
- Nathan Lambert's first open models lecture: distillation, risks and the ecosystem explained — natolambert · 2026-10-26
- Nanbeige 4.3-33B flops real-world test: zero compilable output vs Qwen — julianharris · 2026-10-26
- Claude's safety classifier blocks SS7 defense research, pushing a security dev to an abliterated GLM — l4rz · 2026-10-26