pith. sign in
Pith Number

pith:YURBZWPU

pith:2026:YURBZWPUSUBYO4YGS5VNNET53V
not attested not anchored not stored refs pending

On Cost-Effective LLM-as-a-Judge Improvement Techniques

Luke Markham, Ryan Lail

Ensemble scoring and task-specific criteria injection raise LLM judge accuracy to 85.8 percent on RewardBench 2.

arxiv:2604.13717 v3 · 2026-04-15 · cs.CL

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{YURBZWPUSUBYO4YGS5VNNET53V}

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

Ensemble scoring and task-specific criteria injection together reach up to 85.8% accuracy, +13.5pp over baseline on RewardBench 2.

C2weakest assumption

That the noise-control framing and observed gains on RewardBench 2 will translate to other benchmarks and production use cases without further tuning.

C3one line summary

Ensemble scoring plus task-specific criteria injection raises LLM judge accuracy to 85.8 percent on RewardBench 2, a 13.5-point gain over baseline, with small models gaining the most.

Cited by

1 paper in Pith

Receipt and verification
First computed 2026-06-10T01:10:01.169190Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

c5221cd9f49503877306976ad6927ddd5976e646f53df563ff3824189826ce73

Aliases

arxiv: 2604.13717 · arxiv_version: 2604.13717v3 · doi: 10.48550/arxiv.2604.13717 · pith_short_12: YURBZWPUSUBY · pith_short_16: YURBZWPUSUBYO4YG · pith_short_8: YURBZWPU
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/YURBZWPUSUBYO4YGS5VNNET53V \
  | 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: c5221cd9f49503877306976ad6927ddd5976e646f53df563ff3824189826ce73
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "67a08ea14571d731a88c9854a61cb504b49af35ee55601d57ca2e1395f6aba76",
    "cross_cats_sorted": [],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-04-15T10:52:33Z",
    "title_canon_sha256": "3f0226a141dcee5508df577fbe3c52d73e9c9b666ef1ded637510a61b7bb28f5"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2604.13717",
    "kind": "arxiv",
    "version": 3
  }
}