←
ComputerUseAgents Workshop51 posts
ComputerUseAgents Workshop avatar

ComputerUseAgents Workshop

@workshopcua

Workshop on Computer Use Agents @ ICML 2025 Workshop, Vancouver, Canada

◷ Joined March 2025▧ 13 Photos and videos
ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·19 Jul 2025

Join us at the Computer Use Agents workshop at ICML2025. Happening now in West Meeting Room 211, Vancouver Convention Centre! We have a day packed with fantastic invited and contributed talks, posters, and discussions!

Post attachment
ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·20 May 2025

🧬 HippoRAG: LLM Memory Inspired by the Brain w/ @bernaaaljg et al Yu’s latest work brings neuroscience full circle. HippoRAG, inspired by the hippocampal indexing theory, gives LLMs a robust long-term memory mechanism, essential for persistent, intelligent behavior. #NeuroAI #MemoryInLLMs Don't miss Yu Su at the Computer Use Agents Workshop @ ICML 2025!

ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·20 May 2025

TravelPlanner (ICML 2024 Spotlight) w/ @jianxie_ , @DrogoKhal4 et al A real-world benchmark where language agents tackle complex planning tasks like booking trips—multi-step, goal-oriented, and web-based. #LanguageAgents #WebAutomation 🔗 arxiv.org/abs/2402.01622

ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·19 May 2025

🚨Less than 2 days left to submit your work, check below! Make sure to follow the page and check out our individual speaker spotlights for the #ICML2025 Workshop on Computer Use Agents which will, amongst demo tracks and poster sessions, feature talks and a panel by our amazing speakers: @sercanarik @alexandredrouin @nouhadziri @gneubig @rsalakhu @ysu_nlp @alsuhr @qingyun_wu @zywu_hku @hllo_wrld #ICML2025 #ComputerUse #Agents #WCUA

Post attachment
ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·16 May 2025

📄 RewardBench: Evaluating Reward Models w/ @natolambert, @valentina__py, et al. RewardBench is a benchmark designed to evaluate the capabilities and safety of reward models (including those trained with Direct Preference Optimization, DPO). The dataset is a collection of prompt-chosen-rejected trios spanning chat, reasoning, and safety, to benchmark how reward models perform on challenging, structured and out-of-distribution queries. The authors present many findings on propensity for refusals, reasoning limitations, and instruction following shortcomings of various reward models towards a better understanding of the RLHF process. 🔗 arxiv.org/abs/2403.13787v2

ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·16 May 2025

📄 Faith and Fate:Limits of Transformers on Compositionality w/ @GXiming, @melaniesclar, et al. Under what conditions do Transformers succeed, fail, and why? What types of errors do they make? Can Transformers uncover implicit problem-solving rules or be taught to follow reasoning paths? This work provides novel insights and a nuanced understanding of why and when Transformers succeed or struggle in compositional problems requiring strict multi-hop reasoning to derive correct solutions. To do this it focuses on three tasks: multiplication, logic grid puzzles, and a classic dynamic programming problem. 🔗 arxiv.org/abs/2305.18654

ComputerUseAgents Workshop avatar
ComputerUseAgents Workshop@workshopcua·16 May 2025

📄 WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Models w/ @liweijianglw, @kavel_r, et al. "WildTeaming" is an automatic red-teaming framework that mines in-the-wild user-chatbot interactions to discover 5.7K unique clusters of novel jailbreak tactics, and then composes selections of multiple mined tactics for systematic exploration of novel and even more challenging jailbreaks. WildTeaming intends to address two challenges: (1) 🔍 Broadly identifying jailbroken behaviors of LLMs. (2) 🛠️ Creating a publicly open, large-scale safety training resource for systematic defense (WildJailbreak). 🔗 arxiv.org/abs/2406.18510