Skip to content

Added Rust benchmark suite#302

Open
suleman-dawood wants to merge 5 commits into
mmprotest:mainfrom
suleman-dawood:rust-benchmarks
Open

Added Rust benchmark suite#302
suleman-dawood wants to merge 5 commits into
mmprotest:mainfrom
suleman-dawood:rust-benchmarks

Conversation

@suleman-dawood

Copy link
Copy Markdown
  • Added villani_rust_v1 benchmark suite with 20 Rust tasks (14 bugfix, 3 terminal, 2 localize, 1 two-stage fix)
  • Benchmark result charts for Villani vs Claude Code on Qwen3.5 4B and 9B- 5 easy, 9 medium, 5 hard (+ 1 two-stage fix)
  • UTF-8 slicing, integer truncation, trait bounds, serde roundtrip, negative modulo, sort stability, config precedence, retry policy, cache invalidation, path join, error chains

Results:

  • Villani 4B: 20/20
  • Villani 9B: 20/20
  • Claude Code 4B: 14/20
  • Claude Code 9B: 18/20

Villani achieves 100% solve rate on both model sizes. Claude Code drops tasks due to patch file creation instead of direct edits and missed tool use on some tasks.

> 14 bugfix: chrono, serde, iterators, trait bounds,
  UTF-8 slicing, integer truncation, error chains, two-stage fix
> 3 terminal_workflow: Cargo.toml paths, module structure, features
> 2 localize_patch: feature flag propagation, auth middleware drop
> cargo check passes, cargo test fails on bugs
…ld artifacts which get flagged as forbidden_edit.
…fail with forbidden_edit even

when tests pass
…hentication even when going through a third-party backend service
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant