IBM Research releases ClariEdge dataset and fine-tuned LoRA for query clarification
IBM Research introduced ClariEdge, a new dataset of roughly 3,900 examples that capture ambiguous user queries and negative cases that do not need clarification in retrieval‑augmented generation (RAG) pipelines. The data cover both ill‑formed queries and well‑formed queries that become ambiguous after document retrieval, enabling models to detect when a clarification request is required.
Key points
- ClariEdge dataset contains about 3,900 ambiguous and non‑ambiguous query examples for RAG clarification.
- A lightweight LoRA fine‑tuned model reaches 96.6% classification accuracy, beating prompted larger‑model baselines.
- The same LoRA can be applied before retrieval and after retrieval to filter queries needing clarification.
The authors fine‑tuned a lightweight LoRA adapter on the dataset and reported 96.6% classification accuracy, surpassing prompted baselines of larger language models on both classification and the quality of generated clarification prompts. They highlight that hard negative examples are crucial for distinguishing truly ambiguous queries from those that appear ambiguous but can be resolved from context. The same LoRA can be deployed both before retrieval and after retrieval, acting as a checkpoint in two stages of a RAG system. The ClariEdge dataset is publicly available on Hugging Face at https://huggingface.co/datasets/ibm-research/ClariEdge.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- PHP patches critical RCE flaws discovered with GPT-5.6 Sol, affecting up to 70% of websites · 9 src
- OpenAI and Google AI models discovered critical PHP flaw affecting millions of websites · 8 src
- Researchers release madartb, a parallel Multi-Dialect Arabic dependency treebank · 1 src
- Researchers decode 26 legal factors in three open LLMs · 1 src
- Researchers test how reinforcement learning teaches AI agents to enforce market collusion · 1 src
Comments
via GitHub Discussions