Pith Number

pith:VPZTCO4J

pith:2026:VPZTCO4J4EG4HSUEEIFR2J7HZT

not attested not anchored not stored refs resolved

Teacher-Guided Policy Optimization for LLM Distillation

Bei Li, Chunyang Xiao, Jiahao Liu, Jingang Wang, Jingbo Zhu, Junhao Ruan, Kechen Jiao, Qifan Wang, Runsong Zhao, Tong Xiao, Xin Chen, Xinyu Liu

Teacher-Guided Policy Optimization fixes uninformative feedback in reverse KL by conditioning teacher predictions on student rollouts.

arxiv:2605.13230 v1 · 2026-05-13 · cs.LG · 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{VPZTCO4J4EG4HSUEEIFR2J7HZT}

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

TGPO significantly outperforms standard baselines and is robust to different teachers on complex reasoning benchmarks.

C2weakest assumption

That conditioning teacher predictions on the student's rollout will reliably produce informative directional guidance even when student and teacher distributions diverge substantially.

C3one line summary

TGPO improves on-policy LLM distillation by using teacher predictions conditioned on student rollouts to supply informative guidance when the two distributions diverge.

References

23 extracted · 23 resolved · 18 Pith anchors

[1] SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Models · arXiv:2504.11468

[2] SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training · arXiv:2501.17161

[3] Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge · arXiv:1803.05457

[4] Process Reinforcement through Implicit Rewards · arXiv:2502.01456

[5] MiniLLM: On-Policy Distillation of Large Language Models · arXiv:2306.08543

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-18T02:44:49.594387Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

abf3313b89e10dc3ca84220b1d27e7ccd7906413524cf5e18781579e1ba986ce

Aliases

arxiv: 2605.13230 · arxiv_version: 2605.13230v1 · doi: 10.48550/arxiv.2605.13230 · pith_short_12: VPZTCO4J4EG4 · pith_short_16: VPZTCO4J4EG4HSUE · pith_short_8: VPZTCO4J

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/VPZTCO4J4EG4HSUEEIFR2J7HZT \
  | 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: abf3313b89e10dc3ca84220b1d27e7ccd7906413524cf5e18781579e1ba986ce

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "2c4910f7ce364db0a93c72122ee6e645a59abedf6cfd9fab5f920fd384541c66",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T09:20:03Z",
    "title_canon_sha256": "0e90ea93b8cdaae16215c07b5d8dfcd34c3814bd0bf4212ce3a2310bad428029"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.13230",
    "kind": "arxiv",
    "version": 1
  }
}