daVinci-Dev Agent-native Mid-training for Software Engineering
RESEARCH
-
ConceptMoE: Adaptive Token-to-Concept Compression for Computing
By
–
ConceptMoE Adaptive Token-to-Concept Compression for Implicit Compute Allocation
-

DynamicVLA: Vision-Language-Action Model for Dynamic Object Manipulation
By
–
DynamicVLA
— AK (@_akhaliq) 30 janvier 2026
A Vision-Language-Action Model for Dynamic Object Manipulation pic.twitter.com/wjNl6ASvByDynamicVLA A Vision-Language-Action Model for Dynamic Object Manipulation
-
Scaling Embeddings Outperforms Scaling Experts in Language Models
By
–
Scaling Embeddings Outperforms Scaling Experts in Language Models
-
Benchmarking Spatial Intelligence of Text-to-Image Models
By
–
Everything in Its Place Benchmarking Spatial Intelligence of Text-to-Image Models
-
EU Data Regulations Slow AI Innovation Progress
By
–
AI innovation drops under EU data regulations #AI #AIio #AIInnovation #ML #DataScience #Futureofwork @HaroldSinnott @fogoros @iainljbrown @NandoDF @katecrawford @drhassanrashidi @YuHelenYu
-

ActionMesh: Animated 3D Mesh Generation with Temporal Diffusion
By
–
ActionMesh
— AK (@_akhaliq) 30 janvier 2026
Animated 3D Mesh Generation with Temporal 3D Diffusion pic.twitter.com/WI22IJg3WRActionMesh Animated 3D Mesh Generation with Temporal 3D Diffusion
-
AI Improves Cancer Detection Sensitivity in Radiology Trial
By
–
The false positive rate was 1.4% in both arms of the trial
(radiologist plus AI vs 2 radiologists)
Sensitivity was 80.5% with AI, 73.8% without AI -

Pre-2025 Engineers Bootstrapped Current Generation Models
By
–
Everyone who wrote good code pre-2025 should feel happy that we bootstrapped the current generation of engineers and models
-

Reinforcement Learning via Self-Distillation with Rich Feedback
By
–
"Reinforcement Learning via Self-Distillation" Current RLVR has a major flaw, where credit assignments and signals are sparse due to its binary feedback So this paper introduced a new paradigm called Reinforcement Learning with Rich Feedback (RLRF), using a new