pith. sign in
Pith Number

pith:WZNM7LOH

pith:2026:WZNM7LOHONQGWMACSEN5BJ53ZP
not attested not anchored not stored refs pending

Comparing Developer and LLM Biases in Code Evaluation

Aditya Mittal, Ameet Talwalkar, Chris Donahue, Ryan Shar, Shyam Agarwal, Tongshuang Wu, Valerie Chen, Wayne Chi, Zichu Wu

LLM judges underperform human annotators by 12-23% when predicting developer code preferences across realistic tasks.

arxiv:2603.24586 v2 · 2026-03-25 · cs.SE · cs.CL

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{WZNM7LOHONQGWMACSEN5BJ53ZP}

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

Among 13 different models, the best judges underperform human annotators by 12-23%. TRACE identifies 35 significant sources of misalignment between humans and judges across interaction modalities, the majority of which correspond to existing software engineering code quality criteria.

C2weakest assumption

Human preferences collected via annotation are treated as the ground truth without significant bias or inconsistency, and the automatic rubric extraction accurately isolates the sources of misalignment without introducing artifacts from the extraction process itself.

C3one line summary

TRACE shows LLM judges underperform human annotators by 12-23% and misalign on 35 code quality dimensions across three coding modalities, with biases often matching existing software engineering criteria.

Receipt and verification
First computed 2026-05-17T23:38:59.541346Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

b65acfadc773606b3002911bd0a7bbcbeb6e4e65b1152da91982af76f5399148

Aliases

arxiv: 2603.24586 · arxiv_version: 2603.24586v2 · doi: 10.48550/arxiv.2603.24586 · pith_short_12: WZNM7LOHONQG · pith_short_16: WZNM7LOHONQGWMAC · pith_short_8: WZNM7LOH
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/WZNM7LOHONQGWMACSEN5BJ53ZP \
  | 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: b65acfadc773606b3002911bd0a7bbcbeb6e4e65b1152da91982af76f5399148
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "efa8850c1cee582682bf8b6e383f1415d35e672a8619f1d30c73dc1d426fa080",
    "cross_cats_sorted": [
      "cs.CL"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.SE",
    "submitted_at": "2026-03-25T17:56:55Z",
    "title_canon_sha256": "a5ef48a112ad5f0e5bab1bb3887b3648d3139c5ac89cfd79c1119edae3fe1eab"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2603.24586",
    "kind": "arxiv",
    "version": 2
  }
}