Pith Number

pith:SRP7COU5

pith:2026:SRP7COU5GB3YODBFGUGIILN7V2

not attested not anchored not stored refs resolved

AIS: Adaptive Importance Sampling for Quantized RL

Jiajun Zhou, Lingchao Zheng, Ngai Wong, Wei Shao, Yuwei Fan

Adaptive Importance Sampling corrects non-stationary bias from low-precision rollouts while keeping their speed gains in LLM RL.

arxiv:2605.13907 v1 · 2026-05-13 · stat.ML · 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{SRP7COU5GB3YODBFGUGIILN7V2}

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

AIS matches the BF16 baseline on most tasks while retaining the 1.5 to 2.76x rollout speedup of FP8 by combining weight reliability, divergence severity, and variance amplification into a per-batch mixing coefficient that interpolates between uncorrected and importance-weighted gradients.

C2weakest assumption

The three real-time diagnostics can be combined into a single mixing coefficient that reliably preserves early-training exploration benefits while suppressing later destabilizing bias across different models, tasks, and training stages without introducing new instabilities.

C3one line summary

AIS adaptively corrects non-stationary policy gradient bias in quantized LLM RL, matching BF16 performance while retaining 1.5-2.76x FP8 rollout speedup.

References

33 extracted · 33 resolved · 18 Pith anchors

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

[2] DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models · arXiv:2512.02556

[3] AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning · arXiv:2505.24298

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

[5] OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems · arXiv:2402.14008

Receipt and verification

First computed	2026-05-17T23:39:18.871843Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

945ff13a9d3077870c25350c842dbfaeaae04965786a3c765793336b0568f0e5

Aliases

arxiv: 2605.13907 · arxiv_version: 2605.13907v1 · doi: 10.48550/arxiv.2605.13907 · pith_short_12: SRP7COU5GB3Y · pith_short_16: SRP7COU5GB3YODBF · pith_short_8: SRP7COU5

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/SRP7COU5GB3YODBFGUGIILN7V2 \
  | 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: 945ff13a9d3077870c25350c842dbfaeaae04965786a3c765793336b0568f0e5

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "64b96b9353f36c4ca59c24e29374fefa83ca3be7632760b715f5993f507986ce",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "stat.ML",
    "submitted_at": "2026-05-13T03:36:57Z",
    "title_canon_sha256": "364ed5714d6e429a86c3f08ba45b900a8118869451a10a9058dc9019f851f158"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.13907",
    "kind": "arxiv",
    "version": 1
  }
}