EvoWeb organizes real-world product features into functional dependency graphs, which represent how later requests rely on earlier functionality.
Coding agents struggle to extend web applications across successive requests
EvoWeb tests agents on persistent codebases and finds that integration failures and runtime errors increase as development progresses.
Independent
David Nukrai · Bar Alon · Nativ Levy · Ido Andrew Atad · Ido Cohen · Itamar Zimerman · +3 more
Research Digest··2 min read
Thread:Process Agent Benchmarks
The authors introduce EvoWeb, a benchmark designed to evaluate coding agents under a common development pattern: repeatedly extending an existing web application rather than implementing one fixed specification.
Why this paper
Independent · Part of Process Agent Benchmarks, now 24 papers
In one line
EvoWeb benchmarks coding agents on evolving web applications with 320 cumulative requests across 32 apps.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§