@aclmeeting @NLP4PosImpact @emnlpmeeting.

🚨New #ACL2025NLP tutorial videos!."Navigating Ethical Challenges in NLP: Hands-on strategies for students & researchers" by @LucianaBenotti Fanny Ducel, Karën Fort, Guido Ivetta, @ZhijingJin @knmnyn Seunghun Lee, Minzhi Li, Margot Mieskes, Adriana Pagano

Join millions who have switched to Grok.

I am truly excited to join my fellow Co-chairs @knmnyn, Aurélie Névéol, and awesome committee members David Adelani, Alvin Grissom II, Seunghun Lee, Margot Mieskes, Adriana Silvina Pagano. Warm thx to my predecessor @LucianaBenotti for her incredible contributions & trust in me❤️.

Humbled to take on my 1st ACL Officer position, as Co-Chair of #ACL Ethics Committee. I look forward to promoting Responsible AI in #NLProc community & ensuring ethical research. In my term, I also hope to bridge the narrative w/ AI Safety and various social considerations of AI.

Kudos to all the leading students @YounwooC @ChanglingXavier @_yongjinny @SimkoSamuel @keenansamway @davidguzman1120 + co-advisors of our research @bschoelkopf @radamihalcea @mrinmayasachan @maxhkw!.@MPI_IS @ETH_en @UofTCompSci @VectorInst @TorontoSRI @michigan_AI @UW @UZH_en.

🎉 Big congrats to our students on 3 #EMNLP2025 papers! 1️⃣Agent-to-Agent ToM: LLM Interlocutor Awareness 2️⃣Improving LLM Safety with Contrastive Rep Learning 3️⃣Are LLMs Consequentialist or Deontological?

Awesome work by our talented student @francescortu and collaborators @DiegoDoimo @albecazzaniga @mrinmayasachan @bschoelkopf from @AreaSciencePark @ETH_en @MPI_IS @UniTrieste @UofTCompSci @VectorInst!.

we encourage followup work to try to interpret (1)more LLM mistakes (e.g., handling temporally changing information), (2) "LLM Post-training" as encouraging one type of mechanism & suppressing the other, etc.Check out our papers and

Our "Competitions of Mechanisms" paper proposes an interesting way to interpret LLM behaviors thru how it handles multiple conflicting mechanisms. E.G., in-context knowledge vs. in-weights knowledge🧐This is an elegant philophical way of thinking --

🔍 More details can be found in our paper and github repo Done with @Tsinghua_Uni @UW @MPI_IS @UofTCompSci @VectorInstitute 🎉.

📈 Analysis shows: Optimal selection ratio ~10-15%, our method selects longer responses with richer reward signals, robust across different difficulty-calculation models. 🧠 This could revolutionize how we scale LLM alignment with limited resources!.

💡Method: Rank preference pairs by DPO reward gaps→Select smallest gaps→Train with high-impact data.📊Amazing Results:.1) Outperforms four strong baselines on four popular preference datasets and two alignment tasks.2) Matches or exceeds full dataset performance in 67.5% cases

🎯 Core Innovation:.1) The first method specifically designed for preference data selection.2) Uses the DPO implicit reward mechanism to identify challenging samples.3) Theoretical Analysis for Smaller reward gaps→harder examples→better training signal.

🚀NEW PAPER: Difficulty-Based Preference Data Selection by DPO Implicit Reward Gap🔥We introduce a novel difficulty-based preference data selection method using DPO implicit reward gaps. Spoiler: Selecting the "hardest to distinguish" examples for max learning efficiency!

Congrats again to our brilliant students @davidguzman1120 @_yongjinny for receiving the "Oral Paper Award" at the #ACL2025NLP Workshop on Research on Agent Language Models (REALM)! Check out how Reasoning LLMs optimize self interests over collective success 📊 in our paper

My big pleasure to see our @NLP4PosImpact Workshop grow and glow! We started in 2021, and aim to create a hub for all #NLProc papers with a good will to help social good❤️We have been reaching big impact at #ACL and #EMNLP conferences. Stay tuned for our next edition next year!.

Really enjoyed the @NLP4PosImpact Workshop at @ACLmeeting #ACL2025NLP! I am glad to give the first version of our "LLMs and Democracy Defense" talk ⬇️ We will also present at the upcoming NLP4Democracy Workshop at #COLM2025 in Montreal. Thx to all.

Congrats on the launch!.

Ruyuan is one of the most amazing workshop co-organizers I’ve ever seen! She’s responsible, organized, and humble. Shiny on her are lots of amazing scholarly attitude, hard work, and great teamwork skills.

RT @Jiarui_Liu_: 📢 PAPER ALERT!.Are her voices being heard in AI research?.Our #ACL2025 @NLP4PosImpact paper uncovers complex gender dispar….