Pith Number

pith:HCZGNBM4

pith:2026:HCZGNBM46XR5VSNUAYO5G7WD77

not attested not anchored not stored refs resolved

Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task

Anne G. E. Collins, Danielle Perszyk, Dave August, Gaia Molinaro

Language models diverge from humans by exploiting single solutions rather than gradually exploring goals in self-directed learning tasks.

arxiv:2603.03295 v2 · 2026-02-06 · cs.CL · cs.AI · cs.CY

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{HCZGNBM46XR5VSNUAYO5G7WD77}

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

Across five models (GPT-5, Gemini 2.5 Pro, Claude Sonnet 4.5, Qwen3 32B, and Centaur), we find substantial divergence from human behavior. While people gradually explore and learn to achieve goals with diversity across individuals, most models exploit a single identified solution or show surprisingly low performance, with distinct patterns across models and little variability across instances of the same model.

C2weakest assumption

That the borrowed cognitive science self-directed learning task validly measures the kind of goal selection preferences that LLMs are being asked to replace in real-world agentic, social, or chat settings.

C3one line summary

LLMs diverge from human goal selection in self-directed learning by exploiting single solutions with low variability across instances.

References

24 extracted · 24 resolved · 9 Pith anchors

[1] V ., Arriaga, R

[2] Concrete Problems in AI Safety · arXiv:1606.06565

[3] Large-scale study of curiosity-driven learning · arXiv:1808.04355

[4] Language models trained on media diets can predict public opinion

[5] X., and Schulz, E

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-18T02:45:05.056482Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

38b266859cf5e3dac9b4061dd37ec3fff6255ae31790508eca3613c9c3c819db

Aliases

arxiv: 2603.03295 · arxiv_version: 2603.03295v2 · doi: 10.48550/arxiv.2603.03295 · pith_short_12: HCZGNBM46XR5 · pith_short_16: HCZGNBM46XR5VSNU · pith_short_8: HCZGNBM4

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/HCZGNBM46XR5VSNUAYO5G7WD77 \
  | 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: 38b266859cf5e3dac9b4061dd37ec3fff6255ae31790508eca3613c9c3c819db

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "e43f229a1e3af2371821755f766146cb2f5a4d1ea6fb72e37d1a0ed0f755fc37",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CY"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-02-06T15:39:54Z",
    "title_canon_sha256": "2102fb93400aef01408779e9bfb4813e156e87ef0e913980ec8e92c1454a7877"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2603.03295",
    "kind": "arxiv",
    "version": 2
  }
}