method · first seen 7 Oct 2026
transcoder
Sparse interpretable replacement for MLP layers that decomposes activations into sparse features.
Mentions per week
PapersCodeDiscussion
Earliest sources
- How Do Agentic LLMs Decide to Call Tools? A Tool-Call Vector Shaped by SuppressionPapers · 7 Oct 2026 · Xijie Gong, Tingxu Han, Jiahao Zhang, Wei Song, Ziqi Ding, Hanqi Yan, Youcheng S…
- Comparing Sparsified MLPs and Transcoders: Many Transcoder Features Are Neurons in DisguisePapers · 17 Oct 2026 · Stepan Shabalin, Mateusz Piotrowski
- The Router Already KnowsPapers · 20 Oct 2026 · Andrew Bistras, Marty Johnson
- Circuit Oracle: Automating Attribution Graph Analysis via Natural-Language QueriesPapers · 30 Oct 2026 · Shariar Kabir, Hong Kiat Tan, Swastik Agrawal, Sai V R Chereddy, Sriram Balasubr…
Latest sources
- Circuit Oracle: Automating Attribution Graph Analysis via Natural-Language QueriesPapers · 30 Oct 2026 · Shariar Kabir, Hong Kiat Tan, Swastik Agrawal, Sai V R Chereddy, Sriram Balasubr…
- The Router Already KnowsPapers · 20 Oct 2026 · Andrew Bistras, Marty Johnson
- Comparing Sparsified MLPs and Transcoders: Many Transcoder Features Are Neurons in DisguisePapers · 17 Oct 2026 · Stepan Shabalin, Mateusz Piotrowski
- How Do Agentic LLMs Decide to Call Tools? A Tool-Call Vector Shaped by SuppressionPapers · 7 Oct 2026 · Xijie Gong, Tingxu Han, Jiahao Zhang, Wei Song, Ziqi Ding, Hanqi Yan, Youcheng S…