←
Hadi Amiri755 posts
Hadi Amiri avatar

Hadi Amiri

@amirieb

Assistant Professor of Computer Science at UMass Lowell. Language + Healthcare. Previously at MIT, Harvard, UMD College Park, NUS, and University of Tehran.

◷ Joined June 2011▧ 46 Photos and videos
Machine Learning (ML) Papers avatar
⟳ Hadi Amiri retweeted
Machine Learning (ML) Papers@Memoirs·Jun 21

Fair Cognitive Impairment Detection Through Unlearning William Nguyen, Jiali Cheng, Hadi Amiri arxiv.org/abs/2606.18571 [𝚌𝚜.𝙻𝙶 𝚌𝚜.𝙲𝙻 𝚌𝚜.𝚂𝙳 𝚎𝚎𝚜𝚜.𝙰𝚂] 💬Interspeech 2026

Hadi Amiri avatar
Hadi Amiri@amirieb·Jul 5

This work led by @jialicheng123 moves machine unlearning toward a more mechanistic understanding of why some examples, features, or concepts are easier or harder to remove. Today at 2pm PTD:

Excited that our #ACL2026 paper with zihengchen, @_cagarwal, @amirieb on Measuring Unlearning Difficulty with MechInterp! Time: Sun July 5, 14:00-15:30, Poster Session B Location: Grand Hall Looking forward to connecting with everyone! aclanthology.org/2026.findin…

sijia.liu avatar
⟳ Hadi Amiri retweeted
sijia.liu@sijialiu17·4 Nov 2025

See our MSU–IBM joint work on reasoning model “unlearning” presented at @emnlpmeeting, showcasing how unlearning can be leveraged to sanitize sensitive reasoning traces and enhance model safety.

🎯 Our EMNLP 2025 Main paper “Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills” goes live soon! Catch us on Wednesday in Suzhou at #EMNLP2025 🇨🇳 🔗 Paper Link: arxiv.org/pdf/2506.12963 🏡 Project Page Link: r2mu.netlify.app/ 🗓 November 5, 11:00–12:30 CST (UTC+8) 📍 Hall C, Section 2, 500-Main 🧍 I won’t be there in person — but feel free to chat with my co-authors! 🧠 The Problem You’ve erased sensitive answers from your LRM. But the reasoning traces, the step-by-step “thoughts” that led there, still remain. Even after unlearning, the model can reconstruct or re-infer forgotten answers through these traces. So the question is: 👉 Can we truly forget reasoning traces, while preserving the model’s reasoning ability? 🎯 Our Solution: R²MU (Reasoning-aware Representation Misdirection for Unlearning) We go beyond answer-level forgetting and target the reasoning process itself. R²MU suppresses sensitive reasoning traces while maintaining general reasoning competence. Through representation misdirection, the model unthinks unsafe reasoning paths, while CoT supervision preserves valid reasoning skills. ⚙️ How it Works 🔄 Unthinking Loss: misaligns hidden representations of sensitive reasoning traces with randomized features. 💡 Reasoning Preservation: uses CoT datasets (like LIMO) to retain problem-solving ability. ✅ R²MU erases reasoning traces — not just answers. ✅ Preserves general reasoning and utility across diverse benchmarks. ✅ Achieves the lowest reasoning-trace leakage (RT-UA ↓) on unlearning benchmark WMDP and LRM safety benchmark STAR-1, while maintaining top reasoning accuracy on AIME, MATH-500, and GPQA. 👥 With amazing collaborators from MSU: @ChongyuFan ,@zyh2022 , @jia_jinghan , and my advisor @sijialiu17 . 🙏 Grateful to our IBM collaborators from @MITIBMLab : @NathalieBaraca1 , Dennis Wei, @p_ram_p.

Post attachment
uclanlp avatar
⟳ Hadi Amiri retweeted
uclanlp@uclanlp·26 Oct 2025

We’ve been running the UCLA NLP Seminar for a while now and realized it’s a waste not to share these amazing talks more broadly. So here’s our YouTube channel now! 🎥 Watch and subscribe to our channel for past and upcoming sessions: 👉 youtube.com/@uclanlp-plus #AI #UCLANLP

Dan Jurafsky avatar
⟳ Hadi Amiri retweeted
Dan Jurafsky@jurafsky·24 Aug 2025

Now that school is starting for lots of folks, it's time for a new release of Speech and Language Processing! Jim and I added all sorts of material for the August 2025 release! With slides to match! Check it out here: web.stanford.edu/~jurafsky/s…

Hadi Amiri avatar
Hadi Amiri@amirieb·2 May 2025

EqualizeIR: Mitigating Linguistic Biases in Retrieval Models with @jialicheng123. We show that IR models are biased by the ling. complexity of queries, and introduce 4 types of weak learners to build robust retrievers. shorturl.at/tsw5D #NLProc #NAACL2025

Jessy Li avatar
⟳ Hadi Amiri retweeted
Jessy Li@jessyjli·19 Apr 2025

Thank you so much @amirieb for hosting, amazing visit 🤗 New papers discussed: QUDsim (discourse similarity): arxiv.org/abs/2504.09373 Information salience in LLMs: arxiv.org/abs/2502.14613

This morning we had the pleasure of learning from @jessyjli about how LLMs model curiosity, discourse coherence, and information salience.👉 ow.ly/fZL250BGU4M #Linguistics #Discourse #NLP #LLMs @KCSciences_UML @UT_Linguistics @UTAustin