Agent

124 briefings

• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%
• Deployment Feedback Speeds Up Inference by 6.98×

View topic →

Multimodal

123 briefings

• 300 Tasks Test Visual Reasoning, Retrieval Runs 12.4× Faster
• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%

View topic →

Evaluation

120 briefings

• 300 Tasks Test Visual Reasoning, Retrieval Runs 12.4× Faster
• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%

View topic →

Training

104 briefings

• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%
• Executable Rubrics Cut Evaluation Latency by Up to 320×

View topic →

Image Gen

98 briefings

• Training Bigger Models Without Paying the Full Price
• Life Agents, 4D Video, and Sparse Domain Updates
• Runtime Control Moves Into the Open

View topic →

Safety

93 briefings

• 300 Tasks Test Visual Reasoning, Retrieval Runs 12.4× Faster
• On-Policy Feedback Makes Training Signals Useful
• Executable Rubrics Cut Evaluation Latency by Up to 320×

View topic →

AI for Science

89 briefings

• 300 Tasks Test Visual Reasoning, Retrieval Runs 12.4× Faster
• On-Policy Feedback Makes Training Signals Useful
• Executable Rubrics Cut Evaluation Latency by Up to 320×

View topic →

Efficiency

80 briefings

• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%
• Deployment Feedback Speeds Up Inference by 6.98×

View topic →

Architecture

76 briefings

• Executable Rubrics Cut Evaluation Latency by Up to 320×
• Deployment Feedback Speeds Up Inference by 6.98×
• 3D World-Building Success Stays Below 60%

View topic →

Video Gen

74 briefings

• Voice RAG’s F1 Drop Widens by 67%
• Thirteen Skills Push Hallucination Detection to 92%
• Life Agents, 4D Video, and Sparse Domain Updates

View topic →

Robotics

73 briefings

• Editing Live Video, Shrinking VLMs to 2.7 Bits
• Training Bigger Models Without Paying the Full Price
• Learning Workflows Lift Design Agents Across Seven Setups

View topic →

Reasoning

63 briefings

• Voice RAG’s F1 Drop Widens by 67%
• Editing Live Video, Shrinking VLMs to 2.7 Bits
• Training Bigger Models Without Paying the Full Price

View topic →

Retrieval

63 briefings

• 300 Tasks Test Visual Reasoning, Retrieval Runs 12.4× Faster
• On-Policy Feedback Makes Training Signals Useful
• Voice RAG’s F1 Drop Widens by 67%

View topic →

Interpretability

60 briefings

• Voice RAG’s F1 Drop Widens by 67%
• Executable Rubrics Cut Evaluation Latency by Up to 320×
• Thirteen Skills Push Hallucination Detection to 92%

View topic →

Code Intelligence

36 briefings

• On-Policy Feedback Makes Training Signals Useful
• A 4B Agent Ties the Giants by Verifying, Not Searching Harder
• The Model as Its Own Teacher, and Why Splitting Crowds Wins

View topic →