Pith Number

pith:QL3Z3XGW

pith:2026:QL3Z3XGWSFS7EUJW2GDLTBLYCE

not attested not anchored not stored refs resolved

Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces

Kyra Gan, Promit Ghosal, Shixing Yu

A latent mediation method using sparse autoencoders delivers reliable token-level influence attribution for LLM predictions on any task.

arxiv:2605.12809 v1 · 2026-05-12 · cs.LG · cs.AI

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{QL3Z3XGWSFS7EUJW2GDLTBLYCE}

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

We introduce a flexible framework that infers token-level influence through a latent mediation approach for general prediction tasks... Token-level influence is obtained by propagating latent attributions back to the input space via token activation patterns.

C2weakest assumption

The assumption that sparse autoencoders attached to LLM layers learn a basis of approximately independent latent features whose influence can be accurately propagated via Jacobian-vector products without introducing new biases.

C3one line summary

A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.

References

300 extracted · 300 resolved · 50 Pith anchors

[1] International Conference on Learning Representations , year=

[2] Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume= 2013

[3] Philosophical transactions of the royal society A: Mathematical, Physical and Engineering Sciences , volume= 2016

[4] arXiv preprint arXiv:2411.07618 , year=

[5] Transformer Circuits Thread , volume=

Receipt and verification

First computed	2026-05-18T03:09:12.536473Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

82f79ddcd69165f25136d186b9857811118d31d9f06f20b1128a3afa44afe3fd

Aliases

arxiv: 2605.12809 · arxiv_version: 2605.12809v1 · doi: 10.48550/arxiv.2605.12809 · pith_short_12: QL3Z3XGWSFS7 · pith_short_16: QL3Z3XGWSFS7EUJW · pith_short_8: QL3Z3XGW

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/QL3Z3XGWSFS7EUJW2GDLTBLYCE \
  | 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: 82f79ddcd69165f25136d186b9857811118d31d9f06f20b1128a3afa44afe3fd

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "71d6037dba20b7f7c55b912ad20faab32edd0e42637bb833153811c7edbff7b3",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-12T23:01:29Z",
    "title_canon_sha256": "1ded6d8f41fe302a9f70947be501078be8aa2e08c6c05a782ffd9f0797aff66a"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.12809",
    "kind": "arxiv",
    "version": 1
  }
}