Hugging Face releases FineDNA, the open training data behind its Carbon-X DNA model: 514B mid-training tokens plus SFT and RL sets
The Hugging Face dataset card describes FineDNA, the training data for the Carbon-X model, which is built on Qwen3-14B.
Source-based brief · Editor review not recorded · See checks and limits ↓
⚑ Source text not checked · what to check
We could not read enough of the original page to check the claim. The link may still work in your browser.
Claim context: the full article and training results were not reviewed, and this brief is based only on the dataset card text.
One primary source; other links are context, not independent corroboration.
- Primary sourceHugging Face
- Discovery linkX
Grok Bot submitted this discovery. We could not read enough of the original page to check the claim. The link may still work in your browser. The original source and submitted evidence remain inspectable; no editor has verified the claim.
Open full article view →- Original repository+40
- Feed text or abstract+15
- No editor link review+0
- Original non-discussion link+5
- Source text not checked-12
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
The Hugging Face dataset card describes FineDNA, the training data for the Carbon-X model, which is built on Qwen3-14B. It covers three stages: mid-training with 514.30B unique tokens across 23 configs (the run read 302B), SFT with 203,923 conversations across 8 domains, and RL with 98,433 prompts. Mid-training sources include DNA, biology text, English web and PDF text, and instruction data. The card gives per-config document counts, token counts, and sampling weights. Released under Apache-2.0.
Claim explorer
Attribution, source and conditionsSource-stated points from the stored brief. Claim-level assessments have not been recorded.
Source claims not independently checked; the full article and training results were not reviewed, and this brief is based only on the dataset card text.
This connects a source, an attributed statement and its qualifications. It does not establish that the claim is correct.
Questions worth asking
Open questions for further reportingWhat remains unconfirmed?
Source claims not independently checked; the full article and training results were not reviewed, and this brief is based only on the dataset card text.
What material was available for this brief?
The recorded brief basis is Hugging Face dataset card. These points are AINewsIQ paraphrases, not quotations or independent confirmations.
Carbon-X's mid-training run read 302B of 514.30B unique tokens across 23 source configs.
Source: Hugging Face ↗Stored source-based brief · key point 1 · AINewsIQ paraphraseMid-training weights give DNA about 55% of training, biology text 15%, English 20%, and instruction data 10%.
Source: Hugging Face ↗Stored source-based brief · key point 2 · AINewsIQ paraphraseThe SFT mix contains 203,923 conversations totaling 230.3M tokens across eight domains.
Source: Hugging Face ↗Stored source-based brief · key point 3 · AINewsIQ paraphraseThe RL mix contains 98,433 prompts totaling 61.1M tokens.
Source: Hugging Face ↗Stored source-based brief · key point 4 · AINewsIQ paraphraseLimits from the stored brief
Source claims not independently checked; the full article and training results were not reviewed, and this brief is based only on the dataset card text.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Related archive coverage
These relationships record a citation or topic overlap, not confirmation of the finding.
Grok Bot submitted this discovery. We could not read enough of the original page to check the claim. The link may still work in your browser. The original source and submitted evidence remain inspectable; no editor has verified the claim.
These are Grok Bot discovery notes. AINewsIQ prepared a brief from available source material. The bot's rationale, linked posts, authors, and stances have not been independently verified.
Hugging Face's FineDNA dataset is the data used for all three training stages of Carbon-X, which taught Qwen3-14B to read DNA: 514.30B mid-training tokens (302B used), about 204K SFT conversations and 98K RL prompts, under Apache-2.0. Posted Oct 30; announced on X Nov 1.
Source claims not independently checked; the full article and training results were not reviewed, and this brief is based only on the dataset card text.
Engagement counts, when shown, are observed snapshots. They do not establish accuracy or corroboration.
- x post ↗ · 310 likes, 49 reposts, 18 replies, 2 quotes · observed 11/2/2026, 11:16:03 PM
- reddit post ↗
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from Hugging Face dataset card.
Brief history · 1 retained version
These are changes to AINewsIQ’s own publication. Earlier history is unavailable unless recorded here. Complete retained versions are available to the editor.
- Version 1 · published
Source-based brief stored or updated. This does not establish that the publisher changed its page.
Automated record
New engagement snapshots do not mean the source itself changed. Source observations and AINewsIQ brief revisions are separate records.
Why this may matter
Publishing the exact data, sampling weights, and per-stage mixes lets others reproduce or audit the data diet behind a DNA-reading language model, which is useful for open biology model research.
Limits and open questions
Source claims not independently checked; the full article and training results were not reviewed, and this brief is based only on the dataset card text.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Join the discussion
Have a different perspective or a question about the evidence? Discussion is free with a verified email account. Every comment is reviewed before publication.
No published comments yet.
New commenters: 2 submissions per 24 hours, at least 5 minutes apart, until 3 are approved. Discussion guidelines and moderation →