Eval suite exposing a failure mode in the Gemini 3.5 Flash model (exception-blindness in tabular reconciliation)
-
Updated
May 28, 2026 - Python
#
terminal-bench-3
Here are 4 public repositories matching this topic...
Terminal-Bench 3 task: reverse-engineer a custom binary log format from a partial spec, samples, and a black-box validator
python benchmark reverse-engineering harbor binary-format llm-evaluation ai-benchmark terminal-bench-3
-
Updated
Jun 28, 2026 - Python
Terminal-Bench 3 task for evaluating agentic log analysis, mixed-format parsing, and deterministic JSON reporting.
-
Updated
Jun 25, 2026 - Python
Smallest readable coding-agent harness that scores on benchmarks. ~970 lines, 59.6% on Terminal-Bench 2.0.
benchmark benchmarks software-engineering software-development benchmark-framework software-testing harness harness-framework harness-engineering harness-design harness-ai terminal-bench-3 harness-testing frontier-model-evaluation
-
Updated
Jul 31, 2026 - Python
Improve this page
Add a description, image, and links to the terminal-bench-3 topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the terminal-bench-3 topic, visit your repo's landing page and select "manage topics."