Metraipolis › archive › dc_360427f852bb
LLMs Can Annotate Attribution Graphs — arXiv:2608.02632
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.02632
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:42:07.390Z
- SHA-256 of the archived text
616b40723cfc787e13c6c59d9e209606bf88a2d8b3e323fc5e7789075a936c22- Stable URL of this page
https://metraipolis.com/documento/dc_360427f852bb
How to cite
LLMs Can Annotate Attribution Graphs — arXiv:2608.02632 — arXiv (en), https://arxiv.org/abs/2608.02632 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_360427f852bb (Metraipolis archive dc_360427f852bb, sha256 616b40723cfc787e13c6c59d9e209606bf88a2d8b3e323fc5e7789075a936c22)
BibTeX
@misc{dc360427f852bb,
title = {LLMs Can Annotate Attribution Graphs — arXiv:2608.02632},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.02632},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_360427f852bb; sha256 616b40723cfc787e13c6c59d9e209606bf88a2d8b3e323fc5e7789075a936c22},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02632. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Circuit tracing is an exciting technique for revealing the internal computation of language models, but it requires a time-intensive manual step of grouping individual features or MLP neurons into supernodes. We present a simple pipeline for automating this step: directly presenting feature descriptions to a language model that groups them into supernodes. Using automated interpretability metrics, we confirm that supernodes generated by our pipeline are as interpretable as those generated by human annotators. On a two-hop Capitals task, our pipeline recovers a supernode corresponding to the intermediate hop in 97 of 100 prompts. Finally, we present a simple proof of concept using our pipeline for open-ended exploration, where we automatically annotate 1000 attribution graphs from Wikipedia prompt completions and then use an LLM judge to flag interesting graphs worth human review. We hope this work demonstrates that even simple automation can produce meaningful attribution graph annotations, motivating further work on automated circuit tracing. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.02632 — 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 2026-10-31T12:31:30.308Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).