Swarnashree

ML MTS @reflection_ai ; Past: ML Research @scale_AI ; Grad @LTIatCMU ;

[Swarnashree @swarnashree_ms]

Can finally show what we’ve been up to this year!! Extremely proud of our small but mighty team..!!

[Reflection @reflection_ai]

Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.

- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch

Full weights release this month.

Learn more about Beam: https://t.co/c3Qx2cpM8G

[Graham Neubig @gneubig]

Impressive first release by Reflection!

Good things come to those who wait, apparently 😀

[Reflection @reflection_ai]

Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.

- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch

Full weights release this month.

Learn more about Beam: https://t.co/c3Qx2cpM8G

[Reflection @reflection_ai]

Reflection is partnering with Shinsegae Group to build a 250-megawatt sovereign AI factory for the Republic of Korea.

Open intelligence. Built on trust between allies. Owned by the nations that need it most.

The future of sovereign AI. Read more in the @WSJ.

[Swarnashree @swarnashree_ms]

This is just the beginning and I’m so excited for what lies ahead!

[Reflection @reflection_ai]

Reflection is partnering with Shinsegae Group to build a 250-megawatt sovereign AI factory for the Republic of Korea.

Open intelligence. Built on trust between allies. Owned by the nations that need it most.

The future of sovereign AI. Read more in the @WSJ.

[Zihao Wang @wzihao12]

@JunkaiZZ, @wzihao12, @ybnbxb, @swarnashree_ms, @jaehwanj6, @victorveitch, @WeiWang1973, @_yunzhong, @vbingliu, and lifeng

[Zihao Wang @wzihao12]

On-policy distillation with reverse KL as reward works great—IF you have access to teacher logits.
But what if you don't? What if you want to distill from multiple teachers?
Our solution: distill teacher guidance into rubrics, then do on-policy RL.
Check out our work: https://t.co/5kVYSSUGOB

[Bing Liu @vbingliu]

🔄RLHF → RLVR → Rubrics → OnlineRubrics

👤 Human feedback = noisy & coarse
🧮 Verifiable rewards = too narrow
📋 Static rubrics = rigid, easy to hack, miss emergent behaviors

💡We introduce OnlineRubrics: elicited rubrics that evolve as models train.
https://t.co/YI6pJ7jfJ1

[Bing Liu @vbingliu]

New @Scale_AI paper!

The culprit behind reward hacking? We trace it to misspecification in high-reward tail.

Our fix: rubric-based rewards to tell “excellent” responses apart from “great.”

The result: Less hacking, stronger post-training!  https://t.co/D6aJkZ8zZE

[Swarnashree @swarnashree_ms]

@sama A lone astronaut exploring a mysterious alien planet with a twilight sky, navigating bioluminescent forests, encountering creatures that are communicating through light

[Swarnashree @swarnashree_ms]

@MistralAI You can setup mistral AI 7B in less than 15mins! if you are interested in trying, check this out:
https://t.co/dgWEgs6MIM