Pith Number

pith:PIJPU4DG

pith:2026:PIJPU4DGAG6HKFENONK3GHUOQG

not attested not anchored not stored refs resolved

Temper and Tilt Lead to SLOP: Reward Hacking Mitigation with Inference-Time Alignment

Jing Liu, Toshiaki Koike-Akino, Ye Wang

Adjusting reference-model temperature generalizes inference-time alignment to ensembles of reward models as a sharpened logarithmic opinion pool whose weights can be calibrated to reduce reward hacking.

arxiv:2605.13537 v1 · 2026-05-13 · cs.LG · cs.AI · 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{PIJPU4DGAG6HKFENONK3GHUOQG}

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 propose an algorithm for calibrating SLOP weight parameters and experimentally demonstrate that it improves robustness while preserving alignment performance.

C2weakest assumption

That the proposed calibration algorithm for SLOP weights generalizes beyond the specific experimental setups and that the temperature adjustment reliably extends the theoretical approximations to ensembles without introducing new instabilities.

C3one line summary

Temperature adjustment on the reference model generalizes inference-time alignment to SLOP ensembles of reward models, with a calibration algorithm that improves robustness to reward hacking while preserving alignment performance.

References

17 extracted · 17 resolved · 10 Pith anchors

[1] Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone · arXiv:2404.14219

[2] Best-of-n through the smoothing lens: KL divergence and regret analysis.arXiv preprint arXiv:2507.05913,

[3] Qwen3-VL Technical Report · arXiv:2511.21631

[4] Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback · arXiv:2204.05862

[5] Training Verifiers to Solve Math Word Problems · arXiv:2110.14168

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-18T02:44:24.102420Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

7a12fa706601bc75148d7355b31e8e81bbdea3e86a96a251d0c1c095d063e000

Aliases

arxiv: 2605.13537 · arxiv_version: 2605.13537v1 · doi: 10.48550/arxiv.2605.13537 · pith_short_12: PIJPU4DGAG6H · pith_short_16: PIJPU4DGAG6HKFEN · pith_short_8: PIJPU4DG

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/PIJPU4DGAG6HKFENONK3GHUOQG \
  | 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: 7a12fa706601bc75148d7355b31e8e81bbdea3e86a96a251d0c1c095d063e000

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "2481e8d665fc0f3b97f22844448dad17070e6ff5a7fbcff3521f9ab6946ed229",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T13:47:06Z",
    "title_canon_sha256": "15301f46178a0c9abbd2cf925adeec0b22941843232ebea12d1371a06a6438c2"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.13537",
    "kind": "arxiv",
    "version": 1
  }
}