Pith Number

pith:FPDZ72QB

pith:2023:FPDZ72QBZQE26KWERZH6FTUZ2N

not attested not anchored not stored refs resolved

Directly Fine-Tuning Diffusion Models on Differentiable Rewards

David J Fleet, Kevin Clark, Kevin Swersky, Paul Vicol

Diffusion models can be fine-tuned directly on differentiable rewards by backpropagating gradients through the full sampling process.

arxiv:2309.17400 v2 · 2023-09-29 · cs.CV · 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{FPDZ72QBZQE26KWERZH6FTUZ2N}

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

it is possible to backpropagate the reward function gradient through the full sampling procedure, and that doing so achieves strong performance on a variety of rewards, outperforming reinforcement learning-based approaches.

C2weakest assumption

The reward function must be differentiable with respect to the generated samples, and the sampling process must allow stable gradient flow without excessive variance or memory issues.

C3one line summary

DRaFT fine-tunes diffusion models by differentiating through sampling to maximize rewards, outperforming RL baselines and improving aesthetics on Stable Diffusion 1.4.

References

38 extracted · 38 resolved · 13 Pith anchors

[1] A General Language Assistant as a Laboratory for Alignment · arXiv:2112.00861

[2] Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback · arXiv:2204.05862

[3] Training Diffusion Models with Reinforcement Learning · arXiv:2305.13301

[4] Training Deep Nets with Sublinear Memory Cost · arXiv:1604.06174

[5] Microsoft COCO Captions: Data Collection and Evaluation Server · arXiv:1504.00325

Cited by

29 papers in Pith

DOLLAR: Few-Step Video Generation via Distillation and Latent Reward Optimization

CMAD: Cooperative Multi-Agent Diffusion via Stochastic Optimal Control

How to Guide Your Flow: Few-Step Alignment via Flow Map Reward Guidance

Embedding-perturbed Exploration Preference Optimization for Flow Models

When Preference Labels Fall Short: Aligning Diffusion Models from Real Data

Receipt and verification

First computed	2026-05-17T23:38:48.404088Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

2bc79fea01cc09af2ac48e4fe2ce99d35659314a765873ddd405927a2aa38e2e

Aliases

arxiv: 2309.17400 · arxiv_version: 2309.17400v2 · doi: 10.48550/arxiv.2309.17400 · pith_short_12: FPDZ72QBZQE2 · pith_short_16: FPDZ72QBZQE26KWE · pith_short_8: FPDZ72QB

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/FPDZ72QBZQE26KWERZH6FTUZ2N \
  | 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: 2bc79fea01cc09af2ac48e4fe2ce99d35659314a765873ddd405927a2aa38e2e

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "587acc25088d331ade6829017f847a4b43d8971b06be5caf41cd8c2f78dc6469",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CV",
    "submitted_at": "2023-09-29T17:01:02Z",
    "title_canon_sha256": "2991c385ae42b382a118f4a82555f5882ffacc8d8a6358214db6212e6f3ac194"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2309.17400",
    "kind": "arxiv",
    "version": 2
  }
}