Pith Number

pith:MMV3S43N

pith:2025:MMV3S43NX2NZS5FKPHO33KGNPX

not attested not anchored not stored refs resolved

Spurious Rewards: Rethinking Training Signals in RLVR

Hannaneh Hajishirzi, Luke Zettlemoyer, Nathan Lambert, Pang Wei Koh, Ranjay Krishna, Rui Xin, Rulin Shao, Scott Geng, Sewon Min, Sewoong Oh, Shuyue Stella Li, Simon Shaolei Du, Yiping Wang, Yulia Tsvetkov

Reinforcement learning with verifiable rewards improves math performance in some models even when rewards are random or spurious.

arxiv:2506.10947 v2 · 2025-06-12 · cs.AI · cs.LG

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{MMV3S43NX2NZS5FKPHO33KGNPX}

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

RLVR training with GRPO improves MATH-500 performance for Qwen2.5-Math-7B by 21.4 percentage points using randomly assigned rewards, nearly matching the 29.1-point gain from ground-truth rewards.

C2weakest assumption

The assumption that the performance gains with spurious rewards are primarily driven by the clipping bias in GRPO amplifying specific pretraining behaviors, rather than other unaccounted factors in the training process or model-specific quirks.

C3one line summary

Spurious rewards in RLVR can produce large gains in mathematical reasoning for certain language models via GRPO's clipping bias amplifying pretraining behaviors like code reasoning.

References

21 extracted · 21 resolved · 3 Pith anchors

[1] Nature645(8081), 633–638 (2025) https://doi.org/10.1038/s41586-025-09422-z 2025 · doi:10.1038/s41586-025-09422-z

[2] 2 OLMo 2 Furious · arXiv:2501.00656

[3] CoRR , volume = 2022

[4] ISBN 9 Coverage, Not Averages Semantic Stratification for Trustworthy Retrieval Evaluation 979-8-89176-288-6 2025 · doi:10.18653/v1/2025.acl-industry

[5] Logic-RL: Unleashing LLM Reasoning with Rule-Based Reinforcement Learning 2025 · arXiv:2502.14768

Formal links

1 machine-checked theorem link

Cited by

30 papers in Pith

PRL: Prompts from Reinforcement Learning

Holder Policy Optimisation

A Model Can Help Itself: Reward-Free Self-Training for LLM Reasoning

What Is Preference Optimization Doing, and Why?

Text-to-SPARQL Generation with Reinforcement Learning: A GRPO-based Approach on DBLP

Receipt and verification

First computed	2026-05-17T23:38:47.721148Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

632bb9736dbe9b9974aa79ddbda8cd7dc09d9c242dc0470ab304df9b8ecd32c5

Aliases

arxiv: 2506.10947 · arxiv_version: 2506.10947v2 · doi: 10.48550/arxiv.2506.10947 · pith_short_12: MMV3S43NX2NZ · pith_short_16: MMV3S43NX2NZS5FK · pith_short_8: MMV3S43N

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/MMV3S43NX2NZS5FKPHO33KGNPX \
  | 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: 632bb9736dbe9b9974aa79ddbda8cd7dc09d9c242dc0470ab304df9b8ecd32c5

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "02836581b64f996450743dd106f6e4e88786e5671e9d6a5c5158087a906901cd",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by-nc-sa/4.0/",
    "primary_cat": "cs.AI",
    "submitted_at": "2025-06-12T17:49:55Z",
    "title_canon_sha256": "54ce252e6e4a74379d16b07187cb77f39b51b53b770d20a1668851db184f1cc0"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2506.10947",
    "kind": "arxiv",
    "version": 2
  }
}