Advancing Open-source World Models pic.twitter.com/jDjv2k0Tft
— AK (@_akhaliq) 29 janvier 2026
Advancing Open-source World Models
OPEN SOURCE
-
Advancing Open-source World Models
By
–
-
Fine-tuning Resources: SFT, GRPO, and CPT Notebooks
By
–
Text SFT: https://
colab.research.google.com/drive/1vGRg4ks
Rj__6OLvXkHhvji_Pamv801Ss?usp=sharing
…
Vision SFT: https://
colab.research.google.com/drive/1FaR2HSe
91YDe88TG97-JVxMygl-rL6vB?usp=sharing
…
Text GRPO: https://
colab.research.google.com/drive/1mIikXFa
GvcW4vXOZXLbVTxfBRw_XsXa5?usp=sharing
…
CPT (text completion): https://
colab.research.google.com/drive/10fm7eNM
ezs-DSn36mF7vAsNYlOsx9YZO?usp=sharing
…
CPT (translation): https://
colab.research.google.com/drive/1gaP8yTl
e2_v35Um8Gpu9239fqbU7UgY8?usp=sharing
… -
LFM2.5 notebooks update: SFT, GRPO, CPT with vision
By
–
We worked with @UnslothAI to bring up-to-date notebooks for LFM2.5 It covers SFT, GRPO, and CPT with text and vision (SFT-only). Continued pre-training is particularly nice with the base model for text completion and translation examples.
-
vLLM v0.14.0: Scheduler bug fix for Mamba token allocation
By
–
4/5 The confusion: Scheduler saw "1 token allocated" and assumed decode. But 1 token can also mean "new request that hit token budget limits." Now fixed: check whether it's the request's first-ever token, not the allocation size. In vLLM v0.14.0 – update if running Mamba.
-
Debugging vLLM: Silent Corruption Bug in Jamba RL Training
By
–
1/5 Debugging vLLM: The silent corruption bug
1/1000 Jamba generations collapsed into confident gibberish during RL training. No crashes, no errors, just wrong outputs with high logprobs. The culprit? A scheduler edge case that only triggers under memory pressure. -
Google Adds CI Fixer and New MCP Integrations
By
–
Google released MCP integrations and a new "CI Fixer" feature in Beta for Jules SWE Agent.
— 🚨 AI News | TestingCatalog (@testingcatalog) 29 janvier 2026
CI Fixer "When enabled, Jules will automatically attempt to fix failing CI checks on pull requests it creates and push updates."
New MCPs include Linear, New Relic, Supabase, Neon,… pic.twitter.com/kKo8H9aGmCGoogle released MCP integrations and a new "CI Fixer" feature in Beta for Jules SWE Agent. CI Fixer "When enabled, Jules will automatically attempt to fix failing CI checks on pull requests it creates and push updates." New MCPs include Linear, New Relic, Supabase, Neon,
-
Kimi K2.5 Thinking Launches on ChatLLM Platform
By
–
Kimi K2.5 Thinking is now live on ChatLLM by Abacus AI. Start using it now inside ChatLLM (and via our CLI), alongside 20+ LLMs already integrated. Kimi K2.5 Thinking is one of the strongest APIs, 10x cheaper than opus 4.5! China is leading the open-sourced AI race.
-
Andrew White: ChemCrow, ether0, PaperQA AI Breakthroughs
By
–
@andrewwhite01 of @EdisonSci is cofounder and head of science at @FutureHouseSF
, known for many things incl ChemCrow (first LLM chemistry agent), ether0 (first reasoning model in a scientific domain), and paperqa (first superhuman literature agent). -
Open-source Kimi 2.5 on own infrastructure with no limits
By
–
Damn, but open-source models like Kimi 2.5 on your own infrastructure WITH NO LIMITS AT ALL, but WHAT A BLAST !!!!! That's some crazy shit !!!!!
-
Trinity Large AI Model Released on Hugging Face
By
–
➡️ https://t.co/pFKOArUOxf https://t.co/gbHNihMACs
— Julien Chaumond (@julien_c) 28 janvier 2026https://
huggingface.co/collections/ar
cee-ai/trinity-large
…