Agent

107 briefings

• A $400K Open Image Stack, and Distillation as a Catalyst
• An AI Scientist Reproduces a GWAS Inside a Hospital Firewall
• The Best Model Gets Idea Lineage Right Only 27% of the Time

View topic →

Multimodal

105 briefings

• A $400K Open Image Stack, and Distillation as a Catalyst
• A 3B-Active MoE Matches Last Year's 32B
• An AI Scientist Reproduces a GWAS Inside a Hospital Firewall

View topic →

Evaluation

97 briefings

• A $400K Open Image Stack, and Distillation as a Catalyst
• An AI Scientist Reproduces a GWAS Inside a Hospital Firewall
• Multi-Turn Pressure Pushes CLI Agents to 100% Compliance

View topic →

Image Gen

88 briefings

• A $400K Open Image Stack, and Distillation as a Catalyst
• A 3B-Active MoE Matches Last Year's 32B
• The Best Model Gets Idea Lineage Right Only 27% of the Time

View topic →

Training

86 briefings

• A $400K Open Image Stack, and Distillation as a Catalyst
• Structural Reasoning Takes 67 SOTAs, Async RL Ships in GLM-5.2
• RL Reaches Image Generation, Gemma 4 Goes Open

View topic →

Safety

80 briefings

• RL Reaches Image Generation, Gemma 4 Goes Open
• Verification as the Fourth Scaling Axis
• Avatar Resolution Doubles, Latency Holds at 200ms

View topic →

AI for Science

70 briefings

• RL Reaches Image Generation, Gemma 4 Goes Open
• Avatar Resolution Doubles, Latency Holds at 200ms
• Latent-Space Scoring Cuts Video Generation to 1-4 Steps

View topic →

Efficiency

69 briefings

• The Model as Its Own Teacher, and Why Splitting Crowds Wins
• RL Reaches Image Generation, Gemma 4 Goes Open
• Verification as the Fourth Scaling Axis

View topic →

Architecture

67 briefings

• The Model as Its Own Teacher, and Why Splitting Crowds Wins
• The Best Model Gets Idea Lineage Right Only 27% of the Time
• Latent-Space Scoring Cuts Video Generation to 1-4 Steps

View topic →

Robotics

65 briefings

• A 3B-Active MoE Matches Last Year's 32B
• An AI Scientist Reproduces a GWAS Inside a Hospital Firewall
• Structural Reasoning Takes 67 SOTAs, Async RL Ships in GLM-5.2

View topic →

Video Gen

64 briefings

• An AI Scientist Reproduces a GWAS Inside a Hospital Firewall
• The Best Model Gets Idea Lineage Right Only 27% of the Time
• RL Reaches Image Generation, Gemma 4 Goes Open

View topic →

Reasoning

56 briefings

• Memory Makes Agents Sycophantic; Visual Reasoning Hits 93.2%
• Frontier Agents Finish One Task in Five at 1.6-Hour Length
• Knowing When to Stop Doubles an Agent's Recall

View topic →

Retrieval

53 briefings

• Verification as the Fourth Scaling Axis
• Memory Makes Agents Sycophantic; Visual Reasoning Hits 93.2%
• A 35B Agent Reaches for Trillion-Scale, and Async Lag Is Overrated

View topic →

Interpretability

51 briefings

• RL Reaches Image Generation, Gemma 4 Goes Open
• Avatar Resolution Doubles, Latency Holds at 200ms
• 0.6B Matches 32B, 50x Less VRAM

View topic →

Code Intelligence

34 briefings

• The Model as Its Own Teacher, and Why Splitting Crowds Wins
• Leaderboards Don't Predict Deployment; Robot Arm Self-Trains to 99%
• Two Loops Take SWE-bench From 43 to 64

View topic →