- RL
- RAG
- LLM
- paper-reading
•
•
•
-
LD-MOLE: Learnable Dynamic Routing for Mixture of LoRA Experts
A principled, differentiable alternative to Top-K routing in MoLE — token-aware, layer-aware dynamic expert allocation via the Sparsegen operator.
-
R1-Searcher vs Search-R1: A Tale of Two Cities in RL-RAG
Two papers with near-identical names, published in the same month of 2025 — both train LLMs to search autonomously via RL, but with very different agendas. One is a general framework; the other is a capability-shaping recipe.
-
Google Gemini updates: Flash 1.5, Gemma 2 and Project Astra
We’re sharing updates across our Gemini family of models and a glimpse of Project Astra, our vision for the future of AI assistants.
-
Displaying External Posts on Your al-folio Blog