Issues

• Clarify that task completion measures syntax and grounded tokens, not executable correctness

  openbenchmarks-labs/web-search-for-coding-agents#17 · Filter by author mrmps opened on Sep 29, 2026

• Preserve model-visible evidence for reproducible score replay

  openbenchmarks-labs/web-search-for-coding-agents#16 · Filter by author mrmps opened on Sep 29, 2026

• Disclose and validate the URL-search substitution in Perplexity fetch

  openbenchmarks-labs/web-search-for-coding-agents#15 · Filter by author mrmps opened on Sep 29, 2026

• Document cache freshness and separate default-policy from freshness-controlled results

  openbenchmarks-labs/web-search-for-coding-agents#14 · Filter by author mrmps opened on Sep 29, 2026

• Standardize or disclose per-provider search evidence budgets

  openbenchmarks-labs/web-search-for-coding-agents#13 · Filter by author mrmps opened on Sep 29, 2026

• Match citation URLs by document identity instead of string prefixes

  openbenchmarks-labs/web-search-for-coding-agents#12 · Filter by author mrmps opened on Sep 29, 2026

• Exclude failed fetches from citation grounding

  openbenchmarks-labs/web-search-for-coding-agents#11 · Filter by author mrmps opened on Sep 29, 2026

• hello

  openbenchmarks-labs/web-search-for-coding-agents#6 · Filter by author shreyash1231 opened on Sep 10, 2026