Pith Number

pith:E25CM7BS

pith:2026:E25CM7BSXEBRLIQ6GCAY5C35OP

not attested not anchored not stored refs pending

Fine-tuning vs. In-context Learning in Large Language Models: A Formal Language Learning Perspective

Bishwamittra Ghosh, Deepak Garg, Evimaria Terzi, Krishna P. Gummadi, Mohammad Aflah Khan, Qinyuan Wu, Soumi Das, Till Speicher

Fine-tuning achieves greater language proficiency than in-context learning on in-distribution generalization in formal languages, with equal out-of-distribution performance and diverging inductive biases at high proficiency.

arxiv:2604.23267 v2 · 2026-04-25 · cs.CL · 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{E25CM7BSXEBRLIQ6GCAY5C35OP}

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

FT has greater language proficiency than ICL on in-distribution generalization, but both perform equally well on out-of-distribution generalization. Their inductive biases, measured by the correlation in string generation probabilities, are similar when both modes partially learn the language but diverge at higher proficiency levels.

C2weakest assumption

That success on the discriminative test in formal languages (higher probability for in-language strings) accurately measures language proficiency differences between FT and ICL in a manner relevant to natural language, with the formal task providing sufficient control and no contamination.

C3one line summary

Fine-tuning shows higher proficiency than in-context learning on in-distribution generalization in formal languages, with equal out-of-distribution performance and diverging inductive biases at high proficiency.

Receipt and verification

First computed	2026-05-20T00:05:45.118460Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

26ba267c32b90315a21e30818e8b7d73eae138f2adfbfae1f06132434fded2c0

Aliases

arxiv: 2604.23267 · arxiv_version: 2604.23267v2 · doi: 10.48550/arxiv.2604.23267 · pith_short_12: E25CM7BSXEBR · pith_short_16: E25CM7BSXEBRLIQ6 · pith_short_8: E25CM7BS

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/E25CM7BSXEBRLIQ6GCAY5C35OP \
  | 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: 26ba267c32b90315a21e30818e8b7d73eae138f2adfbfae1f06132434fded2c0

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "09f2bbf13aadb8100f5406375201f4bd022551e18008b36ef94f95cdee684bca",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-04-25T12:19:25Z",
    "title_canon_sha256": "80c87c7ca39a40c603e0714f178c0aabf4cf8c74e21ae3ce61e822d810081ee4"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2604.23267",
    "kind": "arxiv",
    "version": 2
  }
}