Research Insights
- Meta: The Art of Scaling Reinforcement Learning Compute for LLMs.
- The Geometry of Reasoning: Flowing Logics in Representation Space.
- Nvidia: Scaling Test-Time Compute to Achieve IOI Gold Medal with Open-Weight Models.
- Reasoning with Sampling: Your Base Model is Smarter Than You Think.
- Memento: Fine-tuning LLM Agents without Fine-tuning LLMs (github).
- Continual Learning via Sparse Memory Finetuning.
LLM
- Google adds integration between Gemini and Google Maps, allowing geo-grounded answers.
- OpenAI announces ChatGPT Atlas, an AI-enhanced web browser; currently for macOS.
Video
- Google Veo adds precision editing.
Science
- Google DeepMind are partnering with CFS Energy for fusion research.
- Researcher used an LLM to engineer carbon capture enzymes.
Robots