Pith Number

pith:NG7FBDOS

pith:2026:NG7FBDOSXBEX6QYMQUQPFFJ6XA

not attested not anchored not stored refs pending

TIP: Token Importance in On-Policy Distillation

Alborz Geramifard, Hejian Sang, Ran He, Yuanda Xu, Zhengze Zhou, Zhipeng Wang

Informative tokens in on-policy distillation come from high student entropy positions and low-entropy positions with high teacher divergence where the student is overconfident and wrong.

arxiv:2604.14084 v3 · 2026-04-15 · 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{NG7FBDOSXBEX6QYMQUQPFFJ6XA}

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

informative tokens come from two regions: positions with high student entropy, and positions with low student entropy plus high teacher--student divergence, where the student is overconfident and wrong.

C2weakest assumption

That teacher-student divergence reliably flags cases where the student is factually wrong rather than simply differing in style or valid alternative answers.

C3one line summary

In on-policy distillation, tokens with high student entropy or low entropy plus high teacher divergence provide dense corrective signal, allowing effective training on under 20% of tokens across math and planning tasks.

Cited by

9 papers in Pith

SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation

Learning to Foresee: Unveiling the Unlocking Efficiency of On-Policy Distillation

The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs

Receipt and verification

First computed	2026-05-21T01:05:19.098471Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

69be508dd2b8497f430c8520f2953eb838b95aea2f271644dea7ad5cec2258cc

Aliases

arxiv: 2604.14084 · arxiv_version: 2604.14084v3 · doi: 10.48550/arxiv.2604.14084 · pith_short_12: NG7FBDOSXBEX · pith_short_16: NG7FBDOSXBEX6QYM · pith_short_8: NG7FBDOS

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/NG7FBDOSXBEX6QYMQUQPFFJ6XA \
  | 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: 69be508dd2b8497f430c8520f2953eb838b95aea2f271644dea7ad5cec2258cc

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "0beca53e5358430a4cdb5dad9dd376bdad5df3c93db884986e63c800ef1aa22e",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-04-15T16:58:24Z",
    "title_canon_sha256": "1772ecceee47108b85c5ccaa6091b72facab7f9bcc774cc248b1034a9882d7ff"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2604.14084",
    "kind": "arxiv",
    "version": 3
  }
}