Pith Number

pith:YYYCE7JA

pith:2026:YYYCE7JAXC76Q5PEEXX4PR3CWE

not attested not anchored not stored refs pending

Flow Map Language Models: One-step Language Modeling via Continuous Denoising

Aditi Raghunathan, Chanhyuk Lee, Jaehoon Yoo, Jerry Huang, Jinwoo Kim, Manan Agarwal, Nicholas M. Boffi, Seunghoon Hong, Sheel Shah

Continuous flows over one-hot token embeddings match discrete diffusion quality and enable one-step generation that exceeds eight-step baselines.

arxiv:2602.16813 v3 · 2026-02-18 · cs.CL · cs.AI

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{YYYCE7JAXC76Q5PEEXX4PR3CWE}

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

We build a flow language model (FLM), a continuous flow that matches state-of-the-art discrete diffusion baselines on the One Billion Words (LM1B) and OpenWebText (OWT) datasets. We then distill FLM into a flow map language model (FMLM), whose one-step generation exceeds the 8-step quality of recent few-step discrete diffusion language models.

C2weakest assumption

That a continuous flow defined over one-hot token embeddings can be learned such that the associated flow map preserves discrete token structure and yields high-quality samples without requiring additional discrete constraints or post-hoc corrections.

C3one line summary

Continuous flow language models match discrete diffusion baselines and their distilled one-step flow map versions exceed 8-step discrete diffusion quality on LM1B and OWT.

Formal links

2 machine-checked theorem links

Cited by

9 papers in Pith

Contrastive Distribution Matching for Amortized Sequential Monte Carlo in Discrete Diffusion

DiLaDiff: Distilled Latent-Augmented Diffusion for Language Modeling

Continuous Diffusion Scales Competitively with Discrete Diffusion for Language

Drifting Objectives for Refining Discrete Diffusion Language Models

Sampling from Flow Language Models via Marginal-Conditioned Bridges

Receipt and verification

First computed	2026-05-21T02:04:59.369201Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

c630227d20b8bfe875e425efc7c762b10fa4c5619500e55f5688962c9405e32c

Aliases

arxiv: 2602.16813 · arxiv_version: 2602.16813v3 · doi: 10.48550/arxiv.2602.16813 · pith_short_12: YYYCE7JAXC76 · pith_short_16: YYYCE7JAXC76Q5PE · pith_short_8: YYYCE7JA

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/YYYCE7JAXC76Q5PEEXX4PR3CWE \
  | 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: c630227d20b8bfe875e425efc7c762b10fa4c5619500e55f5688962c9405e32c

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "075546144abb6cae9abd8bc29235b660f8d6fd391bcea3423e6091466e61649e",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-02-18T19:23:07Z",
    "title_canon_sha256": "375548dedff37735ecf8b63591b6730351be992e4f018459d7f0776cf108523f"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2602.16813",
    "kind": "arxiv",
    "version": 3
  }
}