Pith Number

pith:GGYGFLDR

pith:2025:GGYGFLDRBY3ACVP3KIOYFIRHNL

not attested not anchored not stored refs pending

Iterative Compositional Data Generation for Robot Control

Anh-Quan Pham, Dani S. Bassett, Eric Eaton, Jorge Mendez-Mendez, Marcel Hussing, Shubhankar P. Patankar

A diffusion transformer factorizes robotic transitions into semantic components and generates data for unseen task combinations after limited training.

arxiv:2512.10891 v5 · 2025-12-11 · cs.RO · 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{GGYGFLDRBY3ACVP3KIOYFIRHNL}

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

Once trained on a limited subset of tasks, our model can zero-shot generate high-quality transitions from which we can learn control policies for unseen task combinations. Our approach substantially improves zero-shot performance over monolithic and hard-coded compositional baselines, ultimately solving nearly all held-out tasks and demonstrating the emergence of meaningful compositional structure in the learned representations.

C2weakest assumption

The robotic domain possesses a clean compositional structure that can be factorized into robot-, object-, obstacle-, and objective-specific components whose interactions are sufficiently captured by attention for reliable zero-shot generalization to arbitrary unseen combinations.

C3one line summary

A compositional diffusion model generates zero-shot data for unseen robotic task combinations and iteratively improves via RL validation, solving nearly all held-out tasks.

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T01:05:03.797303Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

31b062ac710e360155fb521d82a2276afa39201130188326c6dfbc4ad1c0f427

Aliases

arxiv: 2512.10891 · arxiv_version: 2512.10891v5 · doi: 10.48550/arxiv.2512.10891 · pith_short_12: GGYGFLDRBY3A · pith_short_16: GGYGFLDRBY3ACVP3 · pith_short_8: GGYGFLDR

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/GGYGFLDRBY3ACVP3KIOYFIRHNL \
  | 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: 31b062ac710e360155fb521d82a2276afa39201130188326c6dfbc4ad1c0f427

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "e738704ef46298c2c8c1dcc996f4f33d649ea22e41941c7181f81aef4505bf7a",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.RO",
    "submitted_at": "2025-12-11T18:20:49Z",
    "title_canon_sha256": "86d3da3140c7bd9842e6cf20423d3b28835c7f337fb72857a7c57871c93fabea"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2512.10891",
    "kind": "arxiv",
    "version": 5
  }
}