Daniel Yue Zhang 张越
Head of Agents Research, Scale AI
About me
I lead Agents Research at Scale AI, working on Coding, Computer Use, Long-horizon Tool Use, and Cyber Security. Prior to that, I was Post Training lead at Amazon AGI.
I love Metallica and Anime (My current top picks: DBZ, Claymore, Bleach, Attack on Titan, and One Piece).
News
- Oct 2026RSI-Forge: From Research Papers to Environments for Recursive Self-Improvement
- Oct 2026Humanity's Sixth Sense: Benchmarking Intuitive Visual Reasoning in Multimodal Models
- Sep 2026SWE-Bench Pro V2: A Cleaner, Harder-to-Game Leaderboard
- Sep 2026Evidence-First Agentic Verification
- Sep 2026READY or Not: Reliable Enterprise Agent Deployment
- Sep 2026Who Grades the Graders? Rethinking Verifier Design for Computer Use Agents
- Aug 2026Spine-Branch Coordination for Multi-agent Computer Use
- Aug 2026PropUQ-MAS: Propagation-Aware Uncertainty Quantification for LLM Multi-Agent Systems
- Aug 2026Rubric Dropout: A Simple Way to Mitigate Reward Hacking in Rubric-as-Reward RL
- Aug 2026CliniCARE-Bench: Clinical Calibrated Audit of Medical Reasoning in EHR
- Jul 2026Terminal-Bench 3.0: Harder Tasks for Better Agents
- Jul 2026MCP vs. CLI: Does an AI Agent's Tool Interface Still Matter?
- Jun 2026ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks
- Feb 2026MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers
- Dec 2025Amazon Nova 2: Multimodal Reasoning and Generation Models
Brain dump
Stay tuned.