TRM Genesis RealWorld 15K: 15,000 Fact-Based Q&A Examples with Citations

15,000 fact-based question-answer examples designed for training Tiny Recursive Models (TRMs). Within Us AI created this dataset, where each answer includes citations and a verification tag. The dataset was last updated on January 4, 2026.

Use Cases

• Fine-tuning small language models for factual accuracy based on the verified Q&A structure.
• Benchmarking model performance on truthfulness and citation grounding based on the included fact sources.
• Training constraint-aware models for short-context applications based on the TRM-native prompt design.
• Studying methods for integrating citation and verification mechanisms into model outputs.

Strengths

• Contains 15,000 examples, providing a substantial corpus for training.
• Every answer includes fact_sources (citations) and a verification tag, supporting transparency.
• Prompts are described as optimized for tiny contexts, aligning with a specific model architecture.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other specifics like file formats and license are unknown.
• Data may reflect temporal or source bias inherent to its compilation method.

Provenance

Source

Within Us AI

Collection Method

Likely curated or generated for training Tiny Recursive Models; specific gathering method is not detailed.

Time Range

null

Freshness

Last updated 2026-01-04 00:09:22.

Geography

null