TBench Project: Adversarial Benchmark Tasks for LLM Evaluation

evalixa-ai's TBench Project Dataset contains adversarial benchmark tasks designed to evaluate large language models. The dataset was last updated on 2026-05-01. It is hosted on Hugging Face and focuses on reasoning, optimization, and code-related challenges.

Use Cases

• Stress-testing LLM robustness based on adversarial tasks mentioned in the description
• Benchmarking model reasoning and optimization performance based on the described tasks
• Evaluating code generation and understanding in adversarial contexts based on the dataset's focus

Strengths

• Dataset is explicitly designed for adversarial evaluation, a specific and relevant use case
• Focuses on multiple challenging domains including reasoning, optimization, and code
• Last update timestamp is precisely recorded as 2026-05-01 11:40:57

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, file formats, and license are unknown, which may limit suitability assessment

Provenance

Source

evalixa-ai on Hugging Face

Collection Method

Likely curated or generated for benchmark purposes

Time Range

null

Freshness

Last updated 2026-05-01 11:40:57

Geography

null