Harsh Trivedi
@harsh3vedi
🤖 Building AI agents & interactive environments: 🌍 AppWorld (https://appworld.dev) #NLProc PhD @stonybrooku. Prev: @allen_ai @CILVRatNYU. On 🦋 same handle.
🔥 Autonomous AI Assistants (e.g., #googleio2024, #WWDC24) and coding agents (e.g., #Devin, #SWEAgent) have garnered a lot of attention recently. We can envision coding agents autonomously completing complex day-to-day tasks across apps using APIs on our behalf. But how can we…
🚨 I am on the faculty job market this year 🚨 I will be presenting at #NeurIPS2024 and am happy to chat in-person or digitally! I work on developing AI agents that can collaborate and communicate robustly with us and each other. My work covers 3 key problems👇 1⃣ Multi-agent +…
🧵-1 We are thrilled to release #AgentLab, a new open-source package for developing and evaluating web agents. This builds on the new #BrowserGym package which supports 10 different benchmarks, including #WebArena.
There are still a few days left to submit to the AI & Scientific Discovery Workshop at @naaclmeeting ! Both archival and non-archival (i.e. submitted or published) works that you'd like to present to a highly interested audience welcome. ai-and-scientific-discovery.github.io
We (@peterjansen_ai, @mbodhisattwa, @tusharkhot, @harsh3vedi, @Hoper_Tom, @_DougDowney, @erichorvitz) are excited to announce the 📣1st Workshop on AI & Scientific Discovery Workshop (AISD), co-located with NAACL 2025. 📣 tinyurl.com/aisd25
I'm on the academic job market! I build AI systems that assist humans in complicated tasks (e.g. pattern discovery/automate software development), and focus on cases when their outputs are hard-to-explain or evaluate. I'll be at NeurIPS'24 from 12/10-12/15. Happy to catch up!
Excited to give an oral presentation of our work "Controllable Generation via Locally Constrained Resampling" @ #NeurIPS2024 SafeGenAI TL;DR We fix greedy constrained decoding using an ad hoc LLM approximation that we tractably condition on the constraint and reweighing samples
I always believe speech will be the default communication channel between humans and AI agents b/c talking is more efficient, and can convey way more information than text along. Can current audio LMs unlock this potential? To study this, we are launching a new platform Talk…
We (@peterjansen_ai, @mbodhisattwa, @tusharkhot, @harsh3vedi, @Hoper_Tom, @_DougDowney, @erichorvitz) are excited to announce the 📣1st Workshop on AI & Scientific Discovery Workshop (AISD), co-located with NAACL 2025. 📣 tinyurl.com/aisd25
I'm on the faculty market and at #NeurIPS!👩🏫 homes.cs.washington.edu/~niloofar/ I work on privacy, memorization, and emerging challenges in data use for AI. Privacy isn't about PII removal but about controlling the flow of information contextually, & LLMs are still really bad at this!
I’ll be at Johns Hopkins next week, talking about privacy, memorization and language models! Join us!
New CS & @jhuclsp seminar on #GenerativeAI with @niloofar_mire on Monday—don’t miss it! Learn more here: cs.jhu.edu/event/cs-clsp-…
🚨 4th edition of the KAUST Rising Stars in AI Symposium Apply here: kaust.edu.sa/en/news/rising… We'll select the best PhD students, postdocs, early career faculty and industry researchers in AI to present their work at KAUST Deadline: December 18 Please share it widely!
❓Is there any Python lib that can trace LLM calls' input/output + cost locally? It needs to work w/ many LLM provider & orchestrator libs w/o any code change 👉I know wandb/weave supports it by patching LLM libs, but I want a 100% local solution that doesn't require an API key
Excited to host the wonderful @mohitban47 this Friday as part of our Distinguished Lecture Series. The broader @AI_SBU community, the #NLProc, and CV groups at @sbucompsc are looking forward to this. p.s. There will be no remote options for the talk unfortunately.
Looking forward to giving this Distinguished Lecture at StonyBrook next week & meeting the several awesome NLP + CV folks there - thanks @b_niranjan + all for the kind invitation 🙂 PS. Excited to give a new talk on "Planning Agents for Collaborative Reasoning and Multimodal…
🚨 Reverse Thinking Makes LLMs Stronger Reasoners We can often reason from a problem to a solution and also in reverse to enhance our overall reasoning. RevThink shows that LLMs can also benefit from reverse thinking 👉 13.53% gains + sample efficiency + strong generalization!…
Reverse Thinking (analogous to scene reconstruction) helps in reasoning. Check out the thread and the paper.
🚨 Reverse Thinking Makes LLMs Stronger Reasoners We can often reason from a problem to a solution and also in reverse to enhance our overall reasoning. RevThink shows that LLMs can also benefit from reverse thinking 👉 13.53% gains + sample efficiency + strong generalization!…
🚨 Happening today in 5 hours at @CohereForAI! 👉 Consider joining, especially, if you are interested in agentic code generation, tool use, digital automation, and careful environment & benchmark creation for language agents!
🚨 Happening next Monday, 2 Dec, @CohereForAI! ✨ 👋 Anyone can join remotely at this link: 👉 cohere.com/events/cohere-… 📅 Upcoming talks: appworld.dev/talks
My amazing coauthor @steph_milani will present our work at @NeurIPSConf!! Stop by to check how to build LLM applications for mental health training.
🇨🇦 Hi! I’m attending my last @NeurIPSConf as a PhD student, presenting Patient-Ψ at a few workshops. I'm on the job market, looking for TT faculty roles & post-docs. DM if you'd like to chat (or invite me to a party 🥳)!
Monday, December 2nd, check out @harsh3vedi and our Geo Regional Asia group for a talk on "AppWorld: Reliable Evaluation of Interactive Agents in a World of Apps and People." 👥