Pith Number

pith:I44TW35N

pith:2025:I44TW35NPR5PJAAZBXKBHS6KEN

not attested not anchored not stored refs pending

Reward Auditor: Inference on Reward Modeling Suitability in Real-World Perturbed Scenarios

Binhong Li, Hui Liu, Jianxiang Zang, Nijia Mo, Qiang Sun, Ruxue Bai, Shiyu Jiang, Yongda Wei

Reward Auditor uses hypothesis testing to detect if reward models have systematic vulnerabilities under real-world perturbations.

arxiv:2512.00920 v5 · 2025-11-30 · cs.CL

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{I44TW35NPR5PJAAZBXKBHS6KEN}

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

Under real-world perturbed scenarios, Reward Auditor quantifies statistical significance and effect size by auditing distribution degradation of RM preference perception confidence. This enables inference of both the certainty and severity of RM vulnerabilities across diverse real-world scenarios.

C2weakest assumption

The chosen real-world perturbations and the definition of suitability as conditional reliability under those perturbations accurately capture the vulnerabilities that matter for safe LLM alignment in deployment.

C3one line summary

Reward Auditor is a statistical auditing framework that infers systematic vulnerabilities in reward models by quantifying distribution degradation of preference perception confidence under real-world perturbations.

Cited by

3 papers in Pith

The Efficiency Frontier: A Unified Framework for Cost-Performance Optimization in LLM Context Management

PlanViz: Evaluating Planning-Oriented Image Generation and Editing for Computer-Use Tasks

BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models

Receipt and verification

First computed	2026-05-20T00:00:29.144435Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

47393b6fad7c7af480190dd413cbca234016c43855b6f69389c34c9d00954930

Aliases

arxiv: 2512.00920 · arxiv_version: 2512.00920v5 · doi: 10.48550/arxiv.2512.00920 · pith_short_12: I44TW35NPR5P · pith_short_16: I44TW35NPR5PJAAZ · pith_short_8: I44TW35N

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/I44TW35NPR5PJAAZBXKBHS6KEN \
  | 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: 47393b6fad7c7af480190dd413cbca234016c43855b6f69389c34c9d00954930

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "34c28eddc3a84ec52591111031a61ace34d1728a5b1e5a47f6dd1f16ec7690d0",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2025-11-30T14:54:12Z",
    "title_canon_sha256": "58a3a4d390f8261f81cba59cb1c194c6484e3151b0a82d6103dee92fb921eba3"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2512.00920",
    "kind": "arxiv",
    "version": 5
  }
}