Metraipolis › archive › dc_2cc49dc6e0ed

From Input to Output: A Flexible Agent for Dual-End Interpretation of Sparse Autoencoder Features — arXiv:2609.35367

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.35367
License (as archived)
CC0-1.0
Archived
2026-10-07T21:53:41.664Z
SHA-256 of the archived text
2235bd8492aac34b1c3df72b160199dc4d5d6a5a201449fdace1bb809cbcfbc1
Stable URL of this page
https://metraipolis.com/documento/dc_2cc49dc6e0ed

How to cite

From Input to Output: A Flexible Agent for Dual-End Interpretation of Sparse Autoencoder Features — arXiv:2609.35367 — arXiv (en), https://arxiv.org/abs/2609.35367 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2cc49dc6e0ed (Metraipolis archive dc_2cc49dc6e0ed, sha256 2235bd8492aac34b1c3df72b160199dc4d5d6a5a201449fdace1bb809cbcfbc1)
BibTeX
@misc{dc2cc49dc6e0ed,
  title = {From Input to Output: A Flexible Agent for Dual-End Interpretation of Sparse Autoencoder Features — arXiv:2609.35367},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.35367},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2cc49dc6e0ed; sha256 2235bd8492aac34b1c3df72b160199dc4d5d6a5a201449fdace1bb809cbcfbc1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35367. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Sparse autoencoders (SAEs) are an important tool for mechanistic interpretability, but interpreting their many features remains challenging. Existing methods characterize input-side activation patterns and output-side intervention effects, yet often leave their functional connection implicit, while input-side evidence collection typically relies on costly large-corpus scans. We introduce functional interpretation, which characterizes an SAE feature as a mapping from its activating input semantics to its output effects under intervention, and present Dual-End Agentic Feature Interpretation (DAFI), an agent that actively gathers evidence and refines input-side, output-side, and functional interpretations through component-specific feedback. Its short-context token probing enables on-demand activation evidence collection without a full corpus scan. On GemmaScope, DAFI improves Input score by 13.1 percentage points over SAGE and Output score by 38.9 points over Token Change, while being substantially more token-efficient than a general-purpose coding agent. Skills distilled from successful refinements raise the held-out joint pass rate from 58.0% to 92.0% and improve both interpretation quality and efficiency when transferred to a new model-SAE setting. Across features with reliable endpoint interpretations, 70.7% exhibit non-equivalent input and output semantics. On AxBench, DAFI also improves steering-feature selection over output-score filtering. Code is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.35367 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T21:27:08.680Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).