LLMs carry coherent internal body maps, precision scales with model size
Sauers_ · x · 2026-11-11
Sauers released an interpretability study with open code (SauersML/body) showing chat models carry coherent internal body representations when talking about being touched. Experiments run on Qwen3-8B (with Qwen3-32B replications) using plain PyTorch hooks on the residual stream — no SAEs, no fine-tuning. The work explores how the body map changes when the body changes (lost hand, wings, robot body), whether it can be steered, and finds map precision scales with model size. Repo also includes egodeath, self-other, and ownership experiments.
Related event: Study Finds Coherent Body Map Inside LLMs, Improving with Scale(2 posts)→
More from Research
- Probing Qwen3-8B: a simple guesser decodes fictional wing anatomy from hidden states at 100% — Sauers_ · 2026-11-11
- Alice & Bob: ~25,000 cat qubits could factor RSA-2048 in 3 days, slashing prior estimates — AryHHAry · 2026-11-11
- Scandinavian Journal of Statistics editors reflect on publishing in the age of generative AI — michaelchchoi · 2026-11-11
- Proving LLM backdoor-safety without revealing weights: ZKML over formal methods — DimitrisPapail · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11