Pith Number

pith:F6QCW7GL

pith:2026:F6QCW7GLFJAMDP5ZIWUITM533H

not attested not anchored not stored refs resolved

PopuLoRA: Co-Evolving LLM Populations for Reasoning Self-Play

Augustine N. Mavor-Parker, Geoffrey Bradway, Lorenz Wolf, Matthew James Sargent, Maxwill Lin, Roger Creus Castanyer

A population of specialized LoRA adapters in asymmetric self-play creates a co-evolutionary arms race that improves LLM reasoning over single-agent baselines.

arxiv:2605.16727 v1 · 2026-05-16 · cs.AI

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{F6QCW7GLFJAMDP5ZIWUITM533H}

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

Despite lower training-time reward, the population mean outperforms the baseline on three code benchmarks (HumanEval+, MBPP+, LiveCodeBench) and seven math benchmarks (AIME 24/25, AMC 23, MATH-500, Minerva, GSM8K, OlympiadBench), and even the weakest member of the population beats the baseline on aggregate.

C2weakest assumption

That cross-evaluation between sub-populations reliably prevents self-calibration and sustains an expanding problem-space arms race rather than allowing the population to converge on a narrow set of solvable problems.

C3one line summary

PopuLoRA shows that co-evolving populations of LoRA adapters through cross-evaluated self-play can outperform compute-matched single-agent baselines on multiple code and math reasoning benchmarks.

References

67 extracted · 67 resolved · 16 Pith anchors

[1] Vinyals, Oriol and Babuschkin, Igor and Czarnecki, Wojciech M. and Mathieu, Micha. Grandmaster Level in. Nature , volume =. 2019 , doi = 2019

[2] International Conference on Learning Representations , year =

[3] DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models · doi:10.48550/arxiv.2402.03300

[4] Proximal Policy Optimization Algorithms 2017 · doi:10.48550/arxiv.1707.06347

[5] Machine Learning , volume = 1992

Formal links

1 machine-checked theorem link

Receipt and verification

First computed	2026-05-20T00:02:38.695064Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

2fa02b7ccb2a40c1bfb945a889b3bbd9c0a801c5b32458052d00a67bb04eb446

Aliases

arxiv: 2605.16727 · arxiv_version: 2605.16727v1 · doi: 10.48550/arxiv.2605.16727 · pith_short_12: F6QCW7GLFJAM · pith_short_16: F6QCW7GLFJAMDP5Z · pith_short_8: F6QCW7GL

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/F6QCW7GLFJAMDP5ZIWUITM533H \
  | 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: 2fa02b7ccb2a40c1bfb945a889b3bbd9c0a801c5b32458052d00a67bb04eb446

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "7853f28c755e217f94a5d2a661ff5d73c45763cba9f4bf99300ba3cf39976d92",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.AI",
    "submitted_at": "2026-05-16T00:29:35Z",
    "title_canon_sha256": "86dbf4174048556332b84ef277c240c97dde3a1c50fbb125074d6e5dc82c2512"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.16727",
    "kind": "arxiv",
    "version": 1
  }
}