Paper:

TutorRL-7B-think: TutorRL-7B: Github:

AI alignment for tutoring🎓 We use full online RL with conversation-level rewards—not just single-turn signals like DPO. Did the student actually learn by the end?.Using GRPO, the model learns real teaching strategies like when to hint or when to correct. Explore models below⤵️.

🔥 Try it now!.Run MathTutorBench locally with your own models or submit them to our leaderboard. Open-source! 👉@ndaheim_ @idohakimi @ Manu Kapur @IGurevych @mrinmayasachan @ETH_AI_Center.

🤔 𝐌𝐨𝐫𝐞 𝐤𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 ≠ 𝐛𝐞𝐭𝐭𝐞𝐫 𝐭𝐞𝐚𝐜𝐡𝐢𝐧𝐠?.Subject expertise does not always correlate with effective teaching; instead, pedagogy and subject knowledge may present a trade-off.

🎯 How do we measure teaching quality?.We train a reward model that scores open-ended teacher responses and accurately distinguishes expert-level from novice teaching.

📚 Teaching is more than just knowing the answer. Our benchmark goes beyond testing solving ability, evaluating three essential teaching skills (expertise, student understanding, pedagogical ability) across seven diverse tasks using a curated collection of datasets and metrics.

🚀 𝐇𝐨𝐰 𝐰𝐞𝐥𝐥 𝐜𝐚𝐧 𝐋𝐋𝐌𝐬 𝐭𝐞𝐚𝐜𝐡?.Evaluating LLMs for education is key to making real progress, yet we lack a reliable and simple benchmark. Introducing 𝐌𝐚𝐭𝐡𝐓𝐮𝐭𝐨𝐫𝐁𝐞𝐧𝐜𝐡—an open-source benchmark designed to assess holistic tutoring capabilities in AI.

I've been recognized as an Outstanding Reviewer for #EMNLP2024 ! 🚀🚀Contributing to the community is always rewarding and every review is an opportunity to learn and grow.

Mistakes are key learning opportunities!🧑‍🎓 Can LLMs help students learn from them through dialog? 💬 While they often struggle to diagnose student errors when generating responses directly, adding a verification step ✅ could make a difference. #EMNLP2024.

Chat with Junling about our work of generating and evaluating the quality of multi-turn teacher-student conversations grounded in textbooks. #ACL2024 #ACL2024NLP

Excited to be at #EMNLP2023 in Singapore to present our paper.

Consider joining fellowship programs at ETH AI Center.

So many interesting ideas and discussions at #LaunchAIXSummit2023 @ETH_AI_Center.

RT @arkrause: Doctoral and Postdoc Fellowships at the @ETH_AI_Center! Applications accepted until November 22 2023. .

Interested in AI for Education and how LLMs to improve education? Come and join us at the #LaunchAIXSummit2023 workshop tomorrow (10:30)! .@ETH_AI_Center.

LLMs get better at reasoning, but can they act as a tutors? Turns out, they're quick to spill out the answers.🤖 Check out the new 🧮MathDial dataset, built with input from teachers and roleplaying students.

Link to the preprint:

RT @wangchunshu: Tired of the output length limit of ChatGPT? Try RecurrentGPT, a language simulacra of the recurrence mechanism in RNNs. Y….

RT @ETH_AI_Center: Celebrating Jakub Macina's remarkable achievement as he secures a coveted spot on Forbes' 30 Under 30 list in Science an….