Metraipolis › archive › dc_dfed90fd5029

Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety — arXiv:2611.08438

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.08438
License (as archived)
CC0-1.0
Archived
2026-11-06T14:42:57.649Z
SHA-256 of the archived text
a7458d8502e7da4e9ade9364b6f6808a42b404d41b2f2bb34b47ffd24ea29d17
Stable URL of this page
https://metraipolis.com/documento/dc_dfed90fd5029

How to cite

Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety — arXiv:2611.08438 — arXiv (en), https://arxiv.org/abs/2611.08438 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dfed90fd5029 (Metraipolis archive dc_dfed90fd5029, sha256 a7458d8502e7da4e9ade9364b6f6808a42b404d41b2f2bb34b47ffd24ea29d17)
BibTeX
@misc{dcdfed90fd5029,
  title = {Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety — arXiv:2611.08438},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.08438},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dfed90fd5029; sha256 a7458d8502e7da4e9ade9364b6f6808a42b404d41b2f2bb34b47ffd24ea29d17},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.08438. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Neural networks are likely to be among the most consequential artefacts in human history. As these systems become more capable and autonomous, their potential to cause societal harm also grows, making AI safety one of the defining challenges of our time. Yet we continue to build increasingly powerful systems without fully understanding how they work, how they will behave in unfamiliar situations, or how to reliably control them. Mechanistic interpretability offers a path towards understanding these systems by studying their internal representations and computations, with the aim of evaluating, predicting, and controlling behaviour. We survey mechanistic interpretability for AI safety in order to answer the question: Does understanding how models work actually make them safer? We place mechanistic interpretability in the broader context of interpretability research and compare its mechanistic evidence with behavioural evaluation. We introduce the foundations of mechanistic interpretability, including features, polysemanticity, transformer computation, and circuits, and survey methods for feature extraction and analysis, circuit discovery and causal verification, explaining intermediate representations in natural language, learning dynamics, and training-data attribution. Beyond surveying existing methods, we propose a unified framework for evaluating mechanistic interpretability claims and characterise six research orientations by their goals, assumptions, and trade-offs. Finally, we examine how mechanistic understanding can support AI safety, including knowledge editing, steering, and monitoring, and assess its limitations, failure modes, open problems, and relationship to behavioural safety measures. We also release a repository of code notebooks with worked examples that reproduce key findings from surveyed methods, allowing readers to build intuition through hands-on experimentation.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2611.08438 — 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 2026-11-06T14:18:42.703Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).