That's a wrap on #COLM2026! Thank you to everyone who stopped by our booth, joined our evening gathering, & talked open research with us.
Catch up on the papers, interviews, & highlights from the week. 👇
@ai2.bsky.social
@ai2.bsky.socialThat's a wrap on #COLM2026! Thank you to everyone who stopped by our booth, joined our evening gathering, & talked open research with us.
Catch up on the papers, interviews, & highlights from the week. 👇
@mariaa.bsky.socialIf you were missing some more direct advice about “what to do” at #COLM2026, this paper might be a place to look!
@uvp.bsky.socialif you're at #colm2026, why not come listen to some fun #tokenization talks?
Continental 1, all day long
tokenization-workshop.github.io/schedule/
@gautamkamath.comIf you are at #COLM2026, be sure to check out our workshop on responsibly enabling data for foundation models -- starting at 8:45 AM in Union Square 17&18 / Fourth Floor.
Featuring posters and contributed talks, as well as amazing plenary speakers 😲😲😲
@sewon-min-mir-rt.selfhosted.socialRT @i_beltagy: Heading back from #colm2026 with a question I heard over and over from researchers and academics: are you still building Olmo models?
I know it has been an up-and-down year for @allen_ai, but I’m surprised this is even a question.
@j-novikova-nlp.bsky.socialRuss Salakhutdinov and I were both at our first #COLM2026, so I asked him what he made of it. It reminded him of NeurIPS ten years ago, when you could get through all the posters! And he had good advice for early-career researchers.
@google.xmirror.botInterested in privacy-preserving synthetic data?
Catch Alex Bie at the @COLM_conf Google booth (#107) today, at 2:00 PM PT for an encore presentation of ContinuousBench, evaluating knowledge transfer and data contamination in DP synthesis.@GoogleDeepMind
Join the conversation at #COLM2026!
@google.xmirror.botToday at 12:00pm, don't miss our live demonstration on EmbeddingGemma 2 at the @COLM_conf Google booth (#107). Connect with Sahil Dua and explore an open multimodal model that unifies text, images, audio, and video representations. #COLM2026 @GoogleDeepMind
@uvp.bsky.socialgotta say that my take-home from #colm2026 so far is that fun nlp research still exists <3
@jacobeisenstein.bsky.socialTAG mentioned! at #colm2026 (by @adinawilliams.bsky.social maybe?)
in context of this really cool paper on compound tokenization arxiv.org/abs/2610.055...
@phoebenlp.bsky.socialfavorite paper of #colm2026 (so far): CoBPE arxiv.org/pdf/2610.055...
combine modifiers with their heads on the fly to create compositional tokens. shorter sequence length, better downstream results
@hagenblix.bsky.social"[LLMs] are no more gods than are markets or bureaucracies"
@himself.bsky.social at #COLM2026
Precisely the parallelisms that our 'Why We Fear AI' takes as its starting points, too
@google.xmirror.botMissed yesterday's demo on adaptive agent environments?
Stop by the #COLM2026 Google booth #107 today at 11:00 AM PT to catch Zifeng Wang presenting EnvHarness — a flexible framework enabling co-evolution between LLM agents and their training environments.
@amazon.science🎤 Can you out-rhyme the AI agent? Stop by the Amazon booth at @colmweb.org for a live AI rap battle demo, plus talks on multi-agent systems and responsible AI for science: https://amzn.to/4jsMYKM #COLM2026
@mariaa.bsky.socialHenry Farrell is giving a talk on culture and AI now at #COLM2026 and it's off to a blazing start. Join in the ballroom!
@amazon.scienceAI-powered smart glasses that guide delivery drivers from van to the customer's doorstep, hands-free. Hear from Amazon applied scientist Yelin Kim on the computer vision and edge AI behind it at @colmweb.org. #COLM2026.
@davidjurgens.bsky.socialOur work at #COLM2026 asks whether we can measure how well models collaborate cheaply and quickly *before* we configure our multi-agent teams. Drawing techniques from behavioral economics, we show that you can measure the cooperative profile of models—and models do differ!
@davidjurgens.bsky.socialThis afternoon at #COLM2026 (Poster 18): Multi-agent systems are great for complex tasks, but successful completion requires that agents "play nicely" with each other. When one agent hoards all the compute or API credits, this leaves the entire task at risk.
@google.xmirror.botToday at 2:00 PM, join Zifeng Wang at the #COLM2026 Google booth (#107) to learn about EnvHarness, a plug-in architecture that dynamically reshapes environment behaviors to improve reinforcement learning and agent adaptability.
@ai2.bsky.socialWhat’s happening at #COLM2026? 👀🎤
Ai2 Comms Lead @kylelwiggers.bsky.social catches up with Senior Director of NLP @nlpnoah.bsky.social to talk agentic models, the next iteration of Olmo, & more!
At COLM? Come say hi at booth 302! 👋