CO-Bench is a benchmark suite featuring 36 real-world Combinatorial Optimization problems drawn from a broad range of domains and complexity levels. The dataset contains the data for the paper 'CO-Bench: Benchmarking Language Model Agents in Algorithm Search for Combinatorial Optimization'. The dataset page was last updated on 2026-01-12.
Use Cases
- Benchmarking agent performance based on the suite of 36 combinatorial optimization problems.
- Evaluating algorithm search strategies across different domains and complexity levels.
- Training or testing language model agents on structured problem-solving tasks.
Strengths
- Includes 36 distinct combinatorial optimization problems.
- Problems are drawn from a broad range of domains.
- Problems span multiple complexity levels.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- CO-Bench
- Freshness
- Last updated 2026-01-12 05:56:55; freshness should be verified.