Debate over 'GPT-6 edits Qwen weights': critic says it's just direction ablation, not weight editing
IanArawjo · x · 2026-11-03
A viral thread claimed GPT-6 Astra directly edited Qwen3.5-0.8B weights (0.016% changed, math up 6x, no training), dubbing it 'Weight-Space Programming.' @RealAstropulse pushes back hard: the model only used existing tooling to locate a problem area and ablate an associated direction — no raw weight editing — calling the oversell intellectually dishonest.
More from Models
- Matt Pocock: Widely panned Opus 4.7-5 was 'SOTA enough' for me on £180/mo CC Ultra — mattpocockuk · 2026-11-03
- Mistral Large 4 builds real-time 3D tracker for 19,000 satellites — MistralAI · 2026-11-03
- German startup's Endeavor 1.1 claims to beat OpenAI and Anthropic, met with grift accusations — teortaxesTex · 2026-11-03
- Report: DeepSeek V5 to be pretrained almost entirely on Huawei Ascend chips, shunning Nvidia — kimmonismus · 2026-11-03
- MIT: AI edits 0.016% of another model's weights, lifting MATH-500 from 7.9% to 48.1% — rohanpaul_ai · 2026-11-03
- Schmidhuber: AI weight editing is 'old hat' from his 1990s fast weight programmers — SchmidhuberAI · 2026-11-03