Pith Number

pith:WKOZH6FM

pith:2026:WKOZH6FMFL5EXNX3ZVA2B6KHN7

not attested not anchored not stored refs resolved

From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning

Haodong Wu, Jiahao Zhang, Lijie Hu, Yongqi Zhang

Recipe search over fixed instruction pools finds better supervised fine-tuning data than instance ranking or full-data training.

arxiv:2605.12944 v1 · 2026-05-13 · cs.LG · cs.CL

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{WKOZH6FMFL5EXNX3ZVA2B6KHN7}

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

AutoSelection achieves the strongest in-distribution reasoning average across three base models, outperforming full-data training, random recipe search, random top-k, and single-operator selectors.

C2weakest assumption

That the cached task-, data-, and model-side signals plus warmup probes reliably predict full SFT performance so that the limited-budget search can locate high-quality recipes without exhaustive evaluation.

C3one line summary

AutoSelection discovers data recipes from a 90K instruction pool that outperform full-data training and other selectors on reasoning tasks for SFT across multiple models.

References

74 extracted · 74 resolved · 7 Pith anchors

[1] Lima: Less is more for alignment.Advances in Neural Information Processing Systems, 36:55006–55021 2023

[2] What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning

[3] Data-juicer: A one-stop data processing system for large language models 2024

[4] LESS: Selecting influential data for targeted instruction tuning 2024

[5] Task-specific data selection for instruction tuning via monosemantic neuronal activations

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-18T03:09:09.566561Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

b29d93f8ac2afa4bb6fbcd41a0f9476fe17228e45a10ee007869efc9a9671789

Aliases

arxiv: 2605.12944 · arxiv_version: 2605.12944v1 · doi: 10.48550/arxiv.2605.12944 · pith_short_12: WKOZH6FMFL5E · pith_short_16: WKOZH6FMFL5EXNX3 · pith_short_8: WKOZH6FM

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/WKOZH6FMFL5EXNX3ZVA2B6KHN7 \
  | 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: b29d93f8ac2afa4bb6fbcd41a0f9476fe17228e45a10ee007869efc9a9671789

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "6de5f8e7fd7aa067b62fd8eca42fe335d3f9dc1bd6425b7cbe03aa9083805b82",
    "cross_cats_sorted": [
      "cs.CL"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T03:27:21Z",
    "title_canon_sha256": "02b4b61ed71babd76d21aeb1fec96b4e49adac594b0a9b294d29174c74224ecb"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.12944",
    "kind": "arxiv",
    "version": 1
  }
}