Transcoders likely find neuron-aligned features too, and also mislead about structure
banburismus_ · x · 2026-10-18
Extending his SAE argument, banburismus expects transcoders to generally find neuron-aligned or few-neuron combination features — and argues they mostly mislead us about network structure as well, citing comparisons between attribution graph decompositions of arithmetic circuits and neuron-group analysis.
Related event: Researcher: SAEs and Transcoders May Mislead Network Interpretation(2 posts)→
More from Research
- 1-gram wireless mesoscope images cortex-wide activity in freely moving mice — KordingLab · 2026-10-23
- Reversible computing veteran poses open problem: universal BARCS primitives from unbiased Josephson junctions — MikePFrank · 2026-10-23
- Arc Institute uses genomic language model to map 376M prokaryotic noncoding sequences into 5.14M clusters — tedgreenwald · 2026-10-23
- The rise of AI-assisted peer review: reviewers are making it into the appendix — abursuc · 2026-10-23
- Anthropic uses Claude to formally verify security properties in 100+ open-source libraries at RISC-V assembly level — AnthropicAI · 2026-10-23
- Researchers test 9 strategies to pick LLM teams, beating random selection by up to 36.1% — shangbinfeng · 2026-10-23