Pith Number

pith:CGTO6CK7

pith:2025:CGTO6CK7GH5LMD5H2FSJPHWQV7

not attested not anchored not stored refs resolved

OpenCodeReasoning: Advancing Data Distillation for Competitive Coding

Aleksander Ficek, Boris Ginsburg, Jocelyn Huang, Sean Narenthiran, Siddhartha Jain, Somshubra Majumdar, Vahid Noroozi, Wasi Uddin Ahmad

Curating a diverse dataset for supervised fine-tuning lets coding models outperform reinforcement learning on competitive benchmarks.

arxiv:2504.01943 v2 · 2025-04-02 · cs.CL

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

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

Our distilled models use only SFT to achieve 61.8% on LiveCodeBench and 24.6% on CodeContests, surpassing alternatives trained with reinforcement learning.

C2weakest assumption

That prioritizing instruction and solution diversity over code-execution filtering produces superior benchmark performance, as concluded from their post-training analysis.

C3one line summary

A new open SFT dataset for reasoning distillation lets coding models hit state-of-the-art scores on LiveCodeBench and CodeContests with supervised fine-tuning alone, outperforming RL-trained baselines.

References

27 extracted · 27 resolved · 10 Pith anchors

[1] Unified pre-training for program understanding and generation 2021

[2] Unified Pre-training for Program Understanding and Generation 2021 · doi:10.18653/v1/2021.naacl-main.211

[3] Program Synthesis with Large Language Models · arXiv:2108.07732

[4] Evaluating Large Language Models Trained on Code 2025 · arXiv:2107.03374

[5] DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning · arXiv:2501.12948

Formal links

2 machine-checked theorem links

Cited by

21 papers in Pith

Don't Pass@k: A Bayesian Framework for Large Language Model Evaluation

CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignment

NVIDIA Nemotron 3: Efficient and Open Intelligence

A Survey of Reinforcement Learning for Large Reasoning Models

How Robustly do LLMs Understand Execution Semantics?

Receipt and verification

First computed	2026-05-17T23:38:13.307892Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

11a6ef095f31fab60fa7d164979ed0afc5ea6a14d3320363a985ac08f569bb44

Aliases

arxiv: 2504.01943 · arxiv_version: 2504.01943v2 · doi: 10.48550/arxiv.2504.01943 · pith_short_12: CGTO6CK7GH5L · pith_short_16: CGTO6CK7GH5LMD5H · pith_short_8: CGTO6CK7

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/CGTO6CK7GH5LMD5H2FSJPHWQV7 \
  | 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: 11a6ef095f31fab60fa7d164979ed0afc5ea6a14d3320363a985ac08f569bb44

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "f7db4dcfc0769cd1fbf77368df90707c829f2c81aca628cd793299b3693a86ce",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2025-04-02T17:50:31Z",
    "title_canon_sha256": "34c6bea33b5a9cb12fd8d3f3014a2aa0d679dff002e6c13132b8b6ce59ae4f67"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2504.01943",
    "kind": "arxiv",
    "version": 2
  }
}