Blaise

In pursuit of open AGI

[Blaise @_BlaiseAI]

We’re happy to support Guanghan and team in their research as they push dLLM RL forward

Check out d2 now

[Guanghan Wang @Guanghan__Wang]

🚀 Introducing d2 — a principled and efficient RL framework for improving reasoning in diffusion language models (DLMs).

RL works well for autoregressive LLMs.
But for DLMs? It’s fundamentally harder.

We show how to do it right. 👇
📖 https://t.co/Kg5GndV3oA
🌐 https://t.co/YAGUAcspsP
💻 https://t.co/sQKqirA1Re

🧵1/12

[Blaise @_BlaiseAI]

@tyler_griggs_ @NovaSkyAI @lmsysorg @_xjdr Definitely

Will shoot you a DM

[Blaise @_BlaiseAI]

We’re excited to announce Blaise — a lab in pursuit of open AGI — and to do so with several open-source releases

[Blaise @_BlaiseAI]

@NovaSkyAI @lmsysorg @_xjdr @Alibaba_Qwen @NVIDIAAI @deepseek_ai The parser we used to generate the rerollout dataset, with a built-in TUI data viewer

https://t.co/ArIOiG2zvD

[Blaise @_BlaiseAI]

@NovaSkyAI @lmsysorg @_xjdr @Alibaba_Qwen Two large-scale synthetic rerollout datasets based on @NVIDIAAI Nemotron-Agentic-Tool-Use-v1 generated with @deepseek_ai V3.2

https://t.co/7ICznqzsLz
https://t.co/QiOhIpWsav