Agent

89 briefings

• One Token per Evidence Cuts Generation Cost 3-10x
• Swap the Arm Without Retraining; VLMs See Both the Duck and the Rabbit
• Streaming Hand-offs Make Multi-Agent Sharper, ZipSplat Splats With 1/6 the Gaussians

View topic →

Multimodal

87 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster
• Swap the Arm Without Retraining; VLMs See Both the Duck and the Rabbit

View topic →

Evaluation

84 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• One Token per Evidence Cuts Generation Cost 3-10x
• dots.tts Hits 54ms First Packet, SWE Agent Self-Evolves Past 50%

View topic →

Training

77 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• Video Models Stumble on Composite Edits, MoE Fails at the Router
• Streaming Hand-offs Make Multi-Agent Sharper, ZipSplat Splats With 1/6 the Gaussians

View topic →

Image Gen

72 briefings

• One Token per Evidence Cuts Generation Cost 3-10x
• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster
• Video Models Stumble on Composite Edits, MoE Fails at the Router

View topic →

Safety

69 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• One Token per Evidence Cuts Generation Cost 3-10x
• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster

View topic →

Efficiency

60 briefings

• One Token per Evidence Cuts Generation Cost 3-10x
• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster
• NVIDIA Packs Five Modalities Into One Set of Weights

View topic →

Architecture

58 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• One Token per Evidence Cuts Generation Cost 3-10x
• Video Models Stumble on Composite Edits, MoE Fails at the Router

View topic →

AI for Science

56 briefings

• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster
• Video Models Stumble on Composite Edits, MoE Fails at the Router
• dots.tts Hits 54ms First Packet, SWE Agent Self-Evolves Past 50%

View topic →

Robotics

51 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• One Token per Evidence Cuts Generation Cost 3-10x
• A 4B Web Agent Catches Up to Closed CUAs on a Few Thousand Trajectories

View topic →

Reasoning

50 briefings

• DeepSeek V4 Cuts KV to 13.5%, Video Memory Runs 10x Faster
• dots.tts Hits 54ms First Packet, SWE Agent Self-Evolves Past 50%
• NVIDIA Packs Five Modalities Into One Set of Weights

View topic →

Video Gen

49 briefings

• Arbor Triples Research Gains; Environments Become the New Scaling Axis
• One Token per Evidence Cuts Generation Cost 3-10x
• dots.tts Hits 54ms First Packet, SWE Agent Self-Evolves Past 50%

View topic →

Retrieval

47 briefings

• Move to See: Top Model Reaches the Target Just 12% of the Time
• LoRA as a Ruler for Memory, Reversed Video as a Free Counterfactual
• Vision Models Start Redesigning How They Output

View topic →

Interpretability

41 briefings

• Streaming Hand-offs Make Multi-Agent Sharper, ZipSplat Splats With 1/6 the Gaussians
• LoRA as a Ruler for Memory, Reversed Video as a Free Counterfactual
• Agents Start Improving Themselves, and Reaching for Fewer Tools

View topic →

Code Intelligence

31 briefings

• Vision Models Start Redesigning How They Output
• $15 Per Paper, Healthcare Agents Cap at 28%
• Soohak Caps Top Models at 30%

View topic →