Commit History
9 days ago
-
Add Georgia Channing to authors
Verified signature7384908
16 days ago
-
Link the Carbon-X, Carbon-3B and Qwen3-14B models
Verified signature4048e8e -
Link the DNA Intelligence Index dataset
Verified signature95e1e72
21 days ago
-
Polish blog grammar and terminology and fix figure numbering
Co-authored-by: Codex <codex@openai.com>679b4e2 -
Remove local demo video caption
Co-authored-by: Codex <codex@openai.com>ed1c095 -
Rename openness section heading
Co-authored-by: Codex <codex@openai.com>f94b2fa -
Group agent, scale and local deployment under applications
Co-authored-by: Codex <codex@openai.com>9b2cda3 -
Fix grammar in introductory performance claim
Co-authored-by: Codex <codex@openai.com>ce72833 -
Specify open language model in introduction
Co-authored-by: Codex <codex@openai.com>949db90 -
Add partner logos from press release to acknowledgements
Co-authored-by: Codex <codex@openai.com>d1f262d -
Use title case for Carbon-X headline
Co-authored-by: Codex <codex@openai.com>4a90b64 -
Lead with natural language meets DNA title and standfirst
Co-authored-by: Codex <codex@openai.com>34c4bb8 -
Use Carbon-X name in blog footer
Co-authored-by: Codex <codex@openai.com>b630f5b -
Remove draft placeholder notice from blog
Co-authored-by: Codex <codex@openai.com>96a8a37 -
Update author display name to Qiuyi Li
Co-authored-by: Codex <codex@openai.com>d9b721c -
Replace TLDR bullets with italic introduction
Co-authored-by: Codex <codex@openai.com>10eba35 -
Match joint collaboration wording in acknowledgements
Co-authored-by: Codex <codex@openai.com>e665e15 -
Expand TIGEM name in acknowledgements
Co-authored-by: Codex <codex@openai.com>0b4632c -
Add acknowledgements for Carbon-X research collaborators
Co-authored-by: Codex <codex@openai.com>61b2f19 -
Remove scrollbar gutters from auto-sized figure embeds
Co-authored-by: Codex <codex@openai.com>3cb0959 -
Open blog external links outside the Space iframe
Co-authored-by: Codex <codex@openai.com>a91bb12 -
Remove superseded post-training draft notes
Co-authored-by: Codex <codex@openai.com>1a97185 -
Summarize RL gains over SFT across DNA benchmarks
Co-authored-by: Codex <codex@openai.com>b58741e -
Add introduction to verifiable-reward reinforcement learning
Co-authored-by: Codex <codex@openai.com>bcc7760 -
Streamline grounded supervision narrative and position SFT figure
Co-authored-by: Codex <codex@openai.com>1808150 -
Clarify midtraining benefit over RL from Qwen3 baseline
Co-authored-by: Codex <codex@openai.com>9112757 -
Connect midtraining gains to RL and grounded supervision
Move the base-versus-midtrained RL comparison beneath the midtraining section and add the agreed transitions. Co-authored-by: Codex <codex@openai.com>0cc9bf5 -
Streamline midtraining narrative and group supporting figures
Lead with the 300B-token progression, move data and ablations into collapsible boxes, and update figure numbering. Co-authored-by: Codex <codex@openai.com>f7f1833 -
Tighten hybrid tokenizer introduction wording
Co-authored-by: Codex <codex@openai.com>372d19f -
Specify added DNA vocabulary in tokenizer introduction
Co-authored-by: Codex <codex@openai.com>d93fcd2 -
Add token efficiency comparison to tokenizer introduction
Co-authored-by: Codex <codex@openai.com>3808ba7 -
Add hybrid tokenizer opening to DNA learning section
Move the tokenizer figure below the new opening while retaining remaining draft material. Co-authored-by: Codex <codex@openai.com>43199e4 -
Add Diego's first name to author byline
Co-authored-by: Codex <codex@openai.com>29596ef -
Place Dana before Maurizio in author byline
Co-authored-by: Codex <codex@openai.com>e5df884 -
Link Qwen3-14B to its Hugging Face model page
Co-authored-by: Codex <codex@openai.com>4606a64 -
Introduce the three challenges behind the Carbon-X training recipe
Co-authored-by: Codex <codex@openai.com>a74621f -
Remove scoring metrics from benchmark descriptions
Co-authored-by: Codex <codex@openai.com>b80d0ea -
Add three research headings beneath training section title
Co-authored-by: Codex <codex@openai.com>b2037a3 -
Update Space title and short description for Carbon-X
Co-authored-by: Codex <codex@openai.com>2e00d0e -
Add Carbon-X sub-agent evaluation section and figure
Co-authored-by: Codex <codex@openai.com>9a5ad39 -
Remove lessons for adapting collapsible box
Co-authored-by: Codex <codex@openai.com>f4bf952 -
Introduce local llama.cpp demo running on a MacBook
Co-authored-by: Codex <codex@openai.com>af85391 -
Embed local Carbon-X demo below data privacy paragraph
Co-authored-by: Codex <codex@openai.com>403bb1d -
Remove duplicated inference cost discussion from openness section
Co-authored-by: Codex <codex@openai.com>e617aba -
Update Carbon-X cost estimate to about 300 dollars
Align both prose estimates and remove the resolved figure alignment note. Co-authored-by: Codex <codex@openai.com>3ad33a7 -
Standardize figure spacing and trim embedded chart padding
Use consistent 48px figure margins and remove redundant embedded wrapper spacing while retaining room for the tokenizer replay control. Co-authored-by: Codex <codex@openai.com>335a366 -
Smooth transition from Figure 1 to benchmark table
Co-authored-by: Codex <codex@openai.com>4f94db6 -
Clarify comparison model wording in Figure 1 caption
Co-authored-by: Codex <codex@openai.com>9d51036 -
Update Figure 1 caption to use intelligence index
Co-authored-by: Codex <codex@openai.com>8a72324 -
Remove domain expert sentence from introduction
Co-authored-by: Codex <codex@openai.com>4824b9e