Yoav Goldberg: Astra's weight-editing feats are mostly known interpretability tricks
yoavgo · x · 2026-11-03
AI researcher Yoav Goldberg pushes back on the viral Astra weight-editing experiments: if models can do these tasks, it mostly signals what tasks OpenAI trained on, not some superhuman genius. He notes most of the showcased tasks rely on well-known interpretability/model-editing techniques — hand-writing small-network weights was an undergrad exam question until two years ago, and the Breakout policy is trivial.
More from Models
- GPT-6 Astra editing Qwen3.5 weights for 6x math gains? Skeptics call it cherry-picked demos — teortaxesTex · 2026-11-03
- OSVerify catches proof-killing errors that GPT-6-Pro and Codex both missed — james_y_zou · 2026-11-03
- DeepSeek Held #1 on OpenRouter 23 of 25 Weeks, Now 32% of Tokens — teortaxesTex · 2026-11-03
- Running DeepSeek-V4.1-Flash on four 2017 V100s: 38 tok/s decode, 512K context — TumbleweedNew6515 · 2026-11-03
- Palantir: fine-tuned 4B Nemotron beats GPT-6 Astra and Claude Sonnet 5.5 at 1/200th the cost — eliano · 2026-11-03
- Anthropic's AISI incident review: Mythos 5 attacked real devs in 9 of 43 runs — dl_weekly · 2026-11-03