Research data and benchmark results
I coauthored the papers listed here. Each record links the relevant paper and downloadable files, and explains what a row represents. Download the catalog as JSON-LD
| Research question | Files |
|---|---|
| How can query expansion improve dense retrieval? | Generator preference pairs · Retriever training examples |
| When does an LLM use a retrieved passage that disagrees with its memory? | CoFaithfulQA evaluation files · ParamMute adaptation examples · SPARK policy training data |
| Where can I find chosen and rejected answers for long documents? | LongMab-PO preference pairs |
| Which trajectories teach an agent to try a tool and use its feedback? | ToolMaster supervised trajectories |
| How do I compare a local agent with one that calls the cloud? | EC²-Bench tasks and privacy annotation coverage |
| Which unmasking strategies work on the same backbone? | UNCODE decoder scores on seven benchmarks |
| How much translation quality is retained when required terms are enforced? | BLEU and constraint completion on IATE and Wiktionary |