Simons Institute Video Watch posted an update
Evaluating factuality in long-form generation
Why it mattersThe talk proposes breaking long-form AI writing into simpler atomic units to evaluate factuality and faithfulness, with examples including long-form question answering and book summarization. It also discusses scaling evaluation for agentic writing pipelines and open problems such as faulty retrieval and verifying complex claims.
Discuss: How might evaluating atomic units help with factuality in long-form AI writing, and what difficulties remain for complex or interpretive claims?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.