Pith Number

pith:UDLFAG2B

pith:2026:UDLFAG2B7A46W3H7LB2RUCD3D4

not attested not anchored not stored refs resolved

Why Do Reasoning Models Lose Coverage? The Role of Data and Forks in the Road

Chandan K Reddy, Khoa D Doan, Nan Zhang, Ngoc-Hieu Nguyen, Parshin Shojaee, Phuc Minh Nguyen, Rui Zhang

Fine-tuning data with ambiguous decision points causes reasoning models to lose coverage.

arxiv:2605.17026 v1 · 2026-05-16 · 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{UDLFAG2B7A46W3H7LB2RUCD3D4}

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

We hypothesize that this behavior is driven by properties of the fine-tuning data, specifically related to decision points or 'forks in the road' scenarios where model faces indecipherable patterns with multiple valid reasoning paths.

C2weakest assumption

The controlled case studies using graph branching and reasoning modes accurately capture the decision-point dynamics present in real fine-tuning datasets for reasoning models.

C3one line summary

Coverage shrinkage after SFT in reasoning models correlates with prevalence of decision-point scenarios in data and can be partially mitigated by targeted data synthesis and diversity-aware decoding.

References

19 extracted · 19 resolved · 1 Pith anchors

[1] Training Verifiers to Solve Math Word Problems 2021 · doi:10.1038/s41586-025-09422-z

[2] Substitutep=l+7 into the target expression, yieldings=l+18

[3] Substitutel=m+5 into the target expression, yieldings=m+23

[4] Substitutem=f+9 into the target expression, yieldings=f+32

[5] Substitutef=g+11 into the target expression, yieldings=g+43

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T00:03:36.598466Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

a0d6501b41f839eb6cff58751a087b1f00e1514d87ce92f330b0564b43cfde37

Aliases

arxiv: 2605.17026 · arxiv_version: 2605.17026v1 · doi: 10.48550/arxiv.2605.17026 · pith_short_12: UDLFAG2B7A46 · pith_short_16: UDLFAG2B7A46W3H7 · pith_short_8: UDLFAG2B

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/UDLFAG2B7A46W3H7LB2RUCD3D4 \
  | 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: a0d6501b41f839eb6cff58751a087b1f00e1514d87ce92f330b0564b43cfde37

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "d14dbf0a3b6d1cbe75b6f97fc24d666061829bdaafd3036adf4687a76ca40ddb",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-16T14:55:12Z",
    "title_canon_sha256": "d7a0d58a5af3dcdca97a65fc5a3285d611baf18f0803c7355c6a7bf183dc665c"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.17026",
    "kind": "arxiv",
    "version": 1
  }
}