Steering one 'linked' AI twin with an emotion concept transfers it to the other model
Sauers_ · x · 2026-10-29
In the 'neural twins' experiment, Qwen3-4B and AFM-4.5B are linked through hidden-state cross-attention without exchanging tokens. Injecting a concept like 'hatred' or 'disappointment' into one twin steers the other's output as well — no training required — suggesting models may naturally read each other's emotion concepts through such a link. The author frames this as a possible route to inter-model empathy.
More from Research
- Persona prompt pins models at ~35% on SIMMER benchmark regardless of Claude or DeepSeek — amplifiedamp · 2026-10-29
- Alfred Menezes releases gentle introduction to lattice-based crypto covering Kyber and Dilithium — matthew_d_green · 2026-10-29
- Google DeepMind, NUH Singapore Start Trusted Tester Trials of AI Co-Clinician — alan_karthi · 2026-10-29
- NPCI Releases TauIndianBankBench for Tool-Using Banking Agents — NPCI · 2026-10-29
- Axiom's agents are formalizing CFSG in 43M lines of code, possibly the largest open-source project ever — SucceededMind · 2026-10-29
- MIT's 'AI telepathy' KV-cache transfer beats text by 40% accuracy at 4x speed, across model vendors — LucaAmb · 2026-10-29