Hugging Face
Models
Datasets
Spaces
Buckets
Docs
Enterprise
Pricing
Website
Tasks
HuggingChat
Collections
Languages
Organizations
Community
Blog
Posts
Daily Papers
Hardware
Learn
Discord
Forum
GitHub
Solutions
Team & Enterprise
Hugging Face PRO
Enterprise Support
Inference Providers
Inference Endpoints
Storage Buckets
Log In
Sign Up
nmitchko
/
deepseek-v4-Flash-CoLaR
Like
0
Text Generation
PyTorch
colar_reasoning_head
colar
reasoning-compression
latent-reasoning
deepseek-v4
chain-of-thought
arxiv:
2505.16552
License:
deepseek
Model card
Files
Files and versions
xet
Community
Copy to bucket
main
deepseek-v4-Flash-CoLaR
Commit History
Aug 6
13:01
docs: the serving addon is released; point at the newer bundled model
nmitchko
Verified signature
575dd3c
Replaces the two "not released yet" notes with links to github.com/nickmitchko/ds4-reasoning-addon and vllm-ds4-sm120, and flags that DeepSeek-V4-Flash-0731-Latent-Reasoning supersedes this repo.
Jul 15
01:06
Refresh README; document sft-c16 (latent-only reasoning, c=16, learned stop)
nmitchko
Verified signature
3129e95
01:03
Add sft-c16: phase6 Fix-B closed-loop SFT head (c=16, format_version 3)
nmitchko
Verified signature
55889a2
Jul 12
16:01
Update README.md
nmitchko
Verified signature
5c66a36
16:00
Update README.md
nmitchko
Verified signature
7311e53
Jul 11
01:17
Add CoLaR reasoning head (SFT + GRPO), model card, and technical report
nmitchko
Verified signature
84b9eca
Jul 10
14:27
initial commit
nmitchko
Verified signature
7f3e50a
