pith. sign in
Pith Number

pith:AMZB2PK2

pith:2025:AMZB2PK2OIIHM4LVTRFSKH5RSI
not attested not anchored not stored refs resolved

Open Problems in Mechanistic Interpretability

Adria Garriga-Alonso, Alejandro Ortega, Arthur Conmy, Atticus Geiger, Bilal Chughtai, Daniel Murfet, David Bau, Eric J. Michaud, Eric Todd, Jack Lindsey, Jeff Wu, Jesse Hoogland, Jessica Rumbelow, Joseph Bloom, Joseph Miller, Joshua Batson, Lee Sharkey, Lucius Bushnaq, Martin Wattenberg, Max Tegmark, Mor Geva, Nandi Schoots, Neel Nanda, Nicholas Goldowsky-Dill, Stefan Heimersheim, Stella Biderman, Stephen Casper, Tom McGrath, William Saunders

Mechanistic interpretability must solve open problems in methods, applications, and socio-technical challenges to achieve its goals of AI assurance and scientific insight.

arxiv:2501.16496 v1 · 2025-01-27 · cs.LG

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{AMZB2PK2OIIHM4LVTRFSKH5RSI}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp
2 Internet Archive
3 Author claim open · sign in to claim
4 Citations open
5 Replications open
Portable graph bundle live · download bundle · merged state
The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Progress in mechanistic interpretability promises greater assurance over AI system behavior and shed light on exciting scientific questions about the nature of intelligence, but many open problems require solutions before these benefits can be realized.

C2weakest assumption

That solving the identified open problems in methods, applications, and socio-technical challenges will directly produce the promised scientific and engineering benefits.

C3one line summary

A review paper that organizes conceptual, practical, and socio-technical open problems in mechanistic interpretability.

References

77 extracted · 77 resolved · 3 Pith anchors

[1] Understanding the role of individual units in a deep neural network 2024 · doi:10.1073/pnas.1907375117
[2] https://distill.pub/2019/activation-atlas 2019 · doi:10.23915/distill.00015
[3] Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals 2023 · doi:10.1162/tacl_a_00359
[4] Probing for semantic evidence of composition by means of simple classification tasks 2009 · doi:10.18653/v1/w16-2524
[5] ISBN 9781450393522 2024 · doi:10.1145/3531146.3533074

Formal links

2 machine-checked theorem links

Cited by

40 papers in Pith

Receipt and verification
First computed 2026-05-17T23:39:22.194980Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

03321d3d5a72107671759c4b251fb1922e88b10314cc0fd577a0fc72e6fa437b

Aliases

arxiv: 2501.16496 · arxiv_version: 2501.16496v1 · doi: 10.48550/arxiv.2501.16496 · pith_short_12: AMZB2PK2OIIH · pith_short_16: AMZB2PK2OIIHM4LV · pith_short_8: AMZB2PK2
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/AMZB2PK2OIIHM4LVTRFSKH5RSI \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 03321d3d5a72107671759c4b251fb1922e88b10314cc0fd577a0fc72e6fa437b
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "12331aa25501eb215b232c8242d7d5d268c34924e775e461378931445601599e",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2025-01-27T20:57:18Z",
    "title_canon_sha256": "b9c6e9cb0c692d000881d31e45732bf54e9bca012ee1962914207934fab15ba2"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2501.16496",
    "kind": "arxiv",
    "version": 1
  }
}