Pinned post
Sinan Ozdemir
Verified account Fireworks@Prof_OZ
AI Education @FireworksAI_HQ 🔥🎆 NLP + Gen AI Expert / LLM whisperer 🤖 AI Author 📚 Founder @Aikylie (acquired) 📈 Fellow @YCombinator
1.6K Following 2K Followers 562 Posts
Latest posts 5
I saw a great post from @nutlope and wanted to take a shot at making my own Jev-like classifier as well on @FireworksAI_HQ.
I'm so happy people are starting to love LLM-based classifiers so anything I can do to keep the hype going, I'm game.
I'm so happy people are starting to love LLM-based classifiers so anything I can do to keep the hype going, I'm game.
View quoted post on X
I'm especially looking at "mixed across multiple harnesses."
Harnesses change pretty often. Teaching a model to adapt across tasks AND harnesses at this scale will really help future proof our models.
Not a new technique, but cool to see it become more and more the default.
Harnesses change pretty often. Teaching a model to adapt across tasks AND harnesses at this scale will really help future proof our models.
Not a new technique, but cool to see it become more and more the default.
Nearly half a year of silence. We spent it studying one problem: how far RL can scale.View quoted post on X
MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.
Streaming the run: mimo.xiaomi.com/rl/
Companies adopting AI walk the same path:
- Rent closed frontier models
- Engineer harnesses around them
- Move to open models for performance & efficiency
- Train logic & skills into weights
@harvey, @cursor_ai, @cognition know this. AI-native companies just walk it faster.
- Rent closed frontier models
- Engineer harnesses around them
- Move to open models for performance & efficiency
- Train logic & skills into weights
@harvey, @cursor_ai, @cognition know this. AI-native companies just walk it faster.
I gotta try this
Introducing Atlas:
The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.
Model the world, move the camera, and simulate space & time.View quoted post on X
