Are you wondering if concept explanations are any good?? Or just interested in how to evaluate them with automated simulatability? Check out Antonin's great new paper

RT @Antonin_Poche: 🚀 Thrilled to share our new paper (the first of my PhD)! How can we compare concept-based #XAI methods in #NLProc? Co…

RT @GoogleAI: Introducing FACTS Grounding, a new benchmark and leaderboard for evaluating the factuality of LLMs. FACTS was developed in co…

RT @dipanjand: Today, we have released a benchmark and a corresponding leaderboard titled "FACTS Grounding" in collaboration with @kaggle.…

RT @kaggle: Introducing FACTS Grounding. A new benchmark we’re launching with @GoogleDeepMind to evaluate LLM’s factual accuracy on over 17…

RT @goldshtn: Today we published FACTS Grounding, a benchmark and leaderboard for evaluating the factuality of LLMs when grounding to the i…

@yoavgo might as well be!

@yoavgo (a.5) there's no reason to believe that our community's accepted, proper peer review processes would have resulted in the "chosen benchmarks" being any better :)

@yoavgo Bio inspired to me generally meant the pop culture understanding of designing a system inspired by a real creature, step back and watch it improve without understanding how it works, as long as you wired it up faithfully enough

RT @RoyiRassin: How diverse are the outputs of text-to-image models and how can we measure that? In our new work, we propose a measure base…

RT @SatyaScribbles: 🚀 Excited to share the research I worked on during my summer internship at @GoogleAI! We developed FRAMES (Factuality,…

How can we target tasks that challenge both dimensions at a high level? This is a difficult problem - we give a nice discussion around this in the paper with some starting suggestions. The paper is partially a call to action for this important problem. Follow-ups welcome! 🫵