User Preference 564K: Conversation Inputs with Structured Preference Rules

564,000 examples for training lightweight preference extraction models. Each example pairs a conversation input with structured JSON output describing user preferences as condition-action rules. The dataset was created by blackhao0426 and was last updated on March 24, 2026.

Use Cases

• Train lightweight preference extraction models based on the 564K conversation-preference pairs.
• Fine-tune models to generate structured JSON outputs from conversational inputs.
• Develop condition-action rule systems for user preference modeling as described in the dataset's purpose.
• Benchmark models for structured information extraction from dialogue, referencing the dataset's use in training pref-extractor-qwen3-0.6b-full-sft.

Strengths

• Contains 564,000 examples for model training.
• Provides structured JSON outputs paired with conversation inputs, as described.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface, author blackhao0426

Collection Method

Likely created for training the pref-extractor-qwen3-0.6b-full-sft model, a core component of the VARS framework.

Freshness

Last updated 2026-03-24 14:51:40; freshness should be verified.