Researcher trains Qwen3-1.7B to obey activation-space instructions, drawing zigzag cat representations
Sauers_ · x · 2026-10-20
A researcher shows you can fine-tune Qwen3-1.7B to follow instructions about its own internals: prompted to raise its internal 'cat' representation on even layers and lower it on odd layers while describing how a bicycle works, the base model barely moves — but the trained model traces an exact zigzag in activation space per layer while still answering normally.
It's essentially a demo of making 'control your own hidden states' an instruction-followable capability.
Related event: Researcher Trains Qwen3's Internal Representations Into a Zigzag Pattern(9 posts)→
More from Research
- New paper: let models reward hack in RL, then disable the parameters at deployment — nptacek · 2026-10-20
- Victor Taelin: smart enough compilers will understand and swap your algorithms — bronzeagepapi · 2026-10-20
- Galbot proposes observational scaling law for physical recursive self-improvement in VLM agents — Galbot · 2026-10-20
- ActionRoPE encodes action as world coordinates for spatially consistent game world models — Zhiyang Deng · 2026-10-20
- UniAVC unifies audio-video generation via progressive training and asymmetric noise scheduling — Kien T. Pham · 2026-10-20
- AHE paper auto-evolves coding-agent harnesses with decoupled files, lifting Terminal-Bench 2 to 77% — morgymcg · 2026-10-20