AI News HubLIVE
原文

Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline

A new method called Self-Verified Distillation (SVD) enables LLMs to self-improve using only unlabeled prompts, without external feedback. The model generates candidate solutions, filters them through a three-stage verification cascade, and trains on the curated data. Experiments on Qwen3 models show significant gains across math, science, and coding benchmarks.

Article intelligence

EngineersAdvanced

Key points

  • SVD uses cycle-consistency, factuality, and correctness checks to filter self-generated solutions.
  • More candidate samples and larger verification budgets yield higher-quality training data.
  • Qwen3-4B improved by +16.7 in math, +11.1 in science, and +8.3 in coding (pass@1).
  • SVD outperforms test-time compute baselines while requiring only a single inference at test time.

Why it matters

This matters because SVD uses cycle-consistency, factuality, and correctness checks to filter self-generated solutions.

Technical impact

May affect model selection, inference cost, product capability, and evaluation benchmarks.

[2605.26132] Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline

[Submitted on 20 May 2026]

Title:Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline

View a PDF of the paper titled Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline, by Tony Lee and 1 other authors

View PDF

Abstract:Can post-trained large language models (LLMs) further improve themselves using only unlabeled prompts, without external teachers or feedback from tools? We study this setting starting only from unlabeled seed questions with no ground-truth solutions, across three reasoning domains: math, science, and coding. We propose Self-Verified Distillation, a simple post-training refinement algorithm in which the model generates candidate solutions to these seed questions, filters them using prompt-based self-verification, and trains on the resulting self-curated dataset. Inspired by the UQ benchmark's use of multiple validators to screen candidate answers to hard unsolved questions, we adapt this validation-based filtering idea to self-training: the model filters its own generated solutions through a three-stage cascade of cycle-consistency, factuality, and correctness checks, accepting a solution only if it passes all stages with unanimous judge votes. We find that sampling more candidate generations and using a larger verification budget during training data construction produces higher-quality self-curated data and, in turn, better reasoning models. We then train Qwen3 models at multiple scales with Self-Verified Distillation and obtain gains across all three domains. For Qwen3-4B, our method improves aggregate held-out pass@1 by +16.7 points in math (AIME26 and HMMT), +11.1 points in science (GPQA Diamond and HLE), and +8.3 points in coding (LCBv5 and LCBv6), with gains also extending to 0.6B and 8B models. Compared to our test-time-only baseline (UQ-TTC), which improves performance by spending extra compute at inference time, Self-Verified Distillation achieves better performance in most settings while requiring only a single inference call at test time.

Subjects:

Computation and Language (cs.CL); Machine Learning (cs.LG)

Cite as: arXiv:2605.26132 [cs.CL]

(or arXiv:2605.26132v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2605.26132

arXiv-issued DOI via DataCite

Submission history

From: Tony Lee [view email] [v1] Wed, 20 May 2026 17:26:10 UTC (2,264 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Self-Verified Distillation: Your Language Model Is Secretly Its Own Synthetic Data Pipeline, by Tony Lee and 1 other authors

View PDF

TeX Source

view license

Current browse context:

cs.CL

new | recent | 2026-05

Change to browse by:

cs cs.LG

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)