Benchmark Radar Day 110: Your Coding Agent Has Never Met a Robot

Day 110 of Benchmark Radar. The scan fetched 1,548 items, published 676, and marked 352 as recommended.

Author: Koutian Wu; GitHub: ktwu01

Benchmark Radar is live. Open the dashboard, star the repository, or read the paper. The Hugging Face dataset provides the data.

Robo-SWE is a new executable issue-resolution dataset with 236 tasks mined from merged pull requests across 28 robotics repositories. It covers Robot Operating System 2 packages, simulators and physics engines, robotics algorithms, and robot-learning frameworks, packaging each task with its original issue text and a reproducible execution context. If you are evaluating coding agents for robotics work, general software benchmarks skip the specialized build systems, simulation dependencies, timing, and hardware-related behavior those repositories demand. Robo-SWE: An Executable Issue-Resolution Dataset Mined from Robot Software Repositories.

Singularity-Bench is a new open-ended benchmark for agents that build AI systems. Its six end-to-end tasks span pre-training, intermediate training, post-training, inference, agent harnesses, and computing-cluster orchestration, and it compares agent implementations against established human-engineered baselines. If you are evaluating autonomous AI development, this moves the target from isolated coding completion to implementing and operating core system components, though six tasks is a thin basis for broad claims. Singularity-Bench: Can Agents Build Agents?.

Why this matters

A general coding score cannot tell you whether an agent survives a ROS 2 build, a simulator dependency, or a timing constraint. And a pass on one function cannot tell you whether an agent can run a training pipeline end to end. The job itself is the unit worth measuring, not the snippet.

Scan coverage

This is a keyword-filtered discovery feed, not a representative survey of AI research. The archived briefing used 188 of 847 captured evidence records. These findings summarize its cited releases; discovery dates do not necessarily equal original publication dates.

Data and briefing: 2026-11-13 archived snapshot.