Train models to determine whether evidence actually supports financial advertising claims, including their figures and qualifying conditions.

Train models to analyze iOS application security evidence and deliver expert-level answers grounded in platform knowledge.

Evaluate how well models interpret lighting control logic and state transitions using lighting ECU source code and state machine annotations.

Examine comparative ad claims with structured evidence, qualifying conditions, and traceable fact-check labels.

Test whether numerical claims in social ads are genuinely supported by citations with explicit metric definitions and conditions.

Train models to complete engineering simulation code for physics fields, materials, and boundary conditions.

Evaluate code verification and root-cause analysis with 5G core protocol codec and information element edge cases.

Compare promotional claims with supporting evidence to help models determine whether offers are credible, eligible, and still valid.

Help models understand workflow creation and modification requests with natural language examples and structured intent annotations.

The code generation benchmark dataset is a key resource for improving the accuracy of code generation models.

The Graphical User Interface (GUI) dataset focuses on interaction videos with smart devices.

The Public Transport Timetable Image Dataset is a high-quality image collection intended for scientific and technological research.

Use synthetic recommendation copy and structured evidence labels to detect when personal experiences are improperly presented as universal facts.

Help models understand stateful stream processing code through operator pipelines, state transitions, and event-time semantics.

Train precise citation coverage and completeness checks with synthetic social ads and structured evidence annotations.

Connect mobile application security issues with applicable remediation documents to evaluate security retrieval and matching.

Train models to turn advertising briefs into structured social media creative that fits directly into advertising workflows.

Train models to trace how cache requests move from keys to destination nodes using source code and structured reasoning annotations.

Find defects and root causes in cache implementations involving TTL, expiration boundaries, and eviction behavior.

Train models to determine whether health and wellness advertising claims are genuinely supported by their cited evidence, including cases of weak support and overgeneralization.