Agent

26 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open

View topic →

Multimodal

23 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open

View topic →

Training

22 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x
• Token Probabilities as Zero-Shot Rewards Hit 0.95 Correlation

View topic →

Safety

22 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open

View topic →

Efficiency

21 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x

View topic →

Evaluation

20 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x

View topic →

Robotics

17 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x

View topic →

Reasoning

15 briefings

• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x
• 74% of Agent Coordination May Be Wasted Effort

View topic →

Image Gen

15 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open

View topic →

Retrieval

15 briefings

• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open
• 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x

View topic →

AI for Science

12 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open

View topic →

Architecture

12 briefings

• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open
• Example Pairs Replace Prompts, Agents Play Favorites

View topic →

Code Intelligence

10 briefings

• TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open
• Token Probabilities as Zero-Shot Rewards Hit 0.95 Correlation
• Adaptive DiT Patches Hit 3x Speedup, Mamba Improves by Subtraction

View topic →

Interpretability

10 briefings

• Latent Reasoning's Gains Aren't From Reasoning
• Token Probabilities as Zero-Shot Rewards Hit 0.95 Correlation
• Agents Score Higher but Fail the Same Way

View topic →

Video Gen

8 briefings

• Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix
• Agents Score Higher but Fail the Same Way
• Example Pairs Replace Prompts, Agents Play Favorites

View topic →