Chao-Chun (Joe) Hsu (@chaochunh) profile picture

ex-Applied Scientist at AWS Agentic AI. PhD from UChicago @chicagohai

Meguro-ku, Tokyo
Joined August 2012
Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)Photo by Chao-Chun (Joe) Hsu (@chaochunh)
Chao-Chun (Joe) Hsu retweeted
Introducing SID-2, a new search model built on compute-scaling swarms. Our swarms find better results than Opus and Astra, while being 8x faster and 20x lower cost. By 2030, AI will write an internet's worth of text every 24 hours. Current search pipelines were built on hand-designed rules, adding more won't scale to meet this incoming volume of data. SID-2 is available on our API today. Try it on your most demanding searches.
296
460
143
5,450
600,064
Chao-Chun (Joe) Hsu retweeted
Hey everyone! I am super excited to share our new research report! 🎉 Scaling Reranked Retrieval 🚀 Thread with more details! 🧵 (1/17)
19
71
14
310
23,542
Chao-Chun (Joe) Hsu retweeted
We're rolling out pplx-decider-v1.3 and pplx-decider-flash-v1.3, the latest version of our open-weight decision models. v1.3 adds locate, a new feature that extracts spans based on type or instructions. Now available in our Decisions API, @huggingface, and @OpenRouter. perplexity.ai/hub/blog/perpl…
37
37
35
421
254,442
Chao-Chun (Joe) Hsu retweeted
🚀 Recruiting Student Researchers (SRs) interested in Multilingual RL! Current student working on RL, LLM post-training, or multilingual LLMs? We'd love to hear from you! Apply: google.com/about/careers/app… CV + interests: docs.google.com/forms/d/e/1F… Reposts appreciated!
3
25
5
158
36,833
Chao-Chun (Joe) Hsu retweeted
【2027年 Google Student Researcher プログラムの募集開始】 Googleの研究開発チームで学生の皆さんが実際の研究プロジェクトに参画する Student Researcher 2027 プログラムの応募受付が開始されました。この募集には、Google DeepMind 東京拠点におけるポジションも含まれています 🎓 GDM Tokyoでは今年、東大から @GoukiMinegishi @urukas4869 、KAISTから Hyunjun Kimと、計3名のStudent Researcherをチームに迎え入れました。フロンティアAI研究に学生の皆さんが直接携わり、チームのメンバーと議論を重ねながら研究成果を上げてくれたことを嬉しく思っています 💡 修士・博士課程に在籍し、機械学習や音声言語処理・AIエージェントなどの分野で世界にインパクトを与える研究に挑戦したい方にとって絶好の機会です。意欲ある学生の皆さんが、GDM Tokyoでの挑戦に応募することを楽しみにしています。ぜひ以下のリンクからご応募ください 🚀 google.com/about/careers/app…
2
44
3
340
29,029
Chao-Chun (Joe) Hsu retweeted
Introducing e-means, an algorithm beating k-means in 10x fewer epochs. With order of magnitude wall-clock speedups, e-means is released in FAISS and needs no hyperparameter tuning beyond k. It is robust to various scales and datasets. Still skeptical ? Just try e-means [1/10]🧵
150
1,234
258
10,234
900,564
Chao-Chun (Joe) Hsu retweeted
Very excited by the new e-means paper from @swann_bessa et al.! Better convergence at a fraction of the training time than conventional Lloyd kmeans. The blogpost has lots of beautiful visualizations/animations of the studies in the paper!
3
1
23
927
Chao-Chun (Joe) Hsu retweeted
SPARSEUP, the first paper from the @Linkup_platform Research team, is out on arXiv! w/ @Cosimo & Pierre It’s the companion to our blogpost and goes deeper on:  • the ablations behind our simple training recipe, with a closer look at logit shift & fanout  • extended comparisons vs @LightOn’s DenseOn, LateOn & SlideOn (cc @antoine_chaffin @AmelieTabatta @raphaelsrty @paulomouraj) More evidence that learned sparse retrieval holds up. More to come! link: arxiv.org/abs/2611.09680
4
14
2
41
2,799
Chao-Chun (Joe) Hsu retweeted
Is there a way to scale a reasoning model’s judgment for searching a massive corpus? We introduce the Amortized Reasoning Index (ARI), a new kind of retriever that updates its weights during search, using gradients derived at inference time from the ranking preferences of a reasoning model. Every session, current agents spend lots of reasoning deciding which search results matter, and then throw all of that judgment away upon compaction. ARI is our attempt to “amortize” that judgment, so that reasoning spent on past queries keeps improving search on future ones. With no relevance labels, ARI takes a 400M cross-encoder that starts out weak and grows its Recall@100 on held-out queries by up to 7×, ending above Qwen3-Embedding-8B on various oblique/reasoning-intensive search tasks. Read our blogpost: dianetc.com/musings/ari/ Preliminary results and discussions are below!
17
54
13
356
44,254
Chao-Chun (Joe) Hsu retweeted
Back then, we released a preview of our new contextual models, offer an answer to the chunking problem Today we’re releasing a new 0.6B model sharing the embedding space of the 9B, a new public benchmark, as well as strong vision results And it comes with a thorough tech report
Introducing pplx-embed-v2-context: open-weight vision-language embedding models in 9B and 0.6B. Both return one int8 vector per document, or one per chunk that sees the whole document. Read the paper, written with @turbopuffer: arxiv.org/abs/2610.38854
7
9
1
82
4,481
We are adding new ModernBert runtime and supporting more encoders (including decision models)!
packed-encoders v0.3.0 is out 🚀 The release adds support for EmbeddingGemma 2 and LFM2/LFM3 backbones, plus support for our first CPU runtime, ModernBERT And @chaochunh joins the project as a maintainer! Small 🧵 on numbers
8
456
Chao-Chun (Joe) Hsu retweeted
packed-encoders v0.3.0 is out 🚀 The release adds support for EmbeddingGemma 2 and LFM2/LFM3 backbones, plus support for our first CPU runtime, ModernBERT And @chaochunh joins the project as a maintainer! Small 🧵 on numbers
1
2
1
8
642
Chao-Chun (Joe) Hsu retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
2,368
7,253
5,759
55,905
10,098,363
Chao-Chun (Joe) Hsu retweeted
Love BrowseComp-Plus? Meet 𝐂𝐌𝐀𝐒𝐒! 🚀 cmass.tech 🔥Your next challenge for agentic search: 423 question–answer pairs spanning BrowseComp-Plus, DeepSearchQA, Total Recall QA, and HotpotQA—all projected onto ClimbMix, a realistic corpus of over 550M documents.
2
8
25
1,023
Chao-Chun (Joe) Hsu retweeted
I improved NanoGPT Speedrun from 39.9s → 6.7s (-83.2%), more than every record since 2025 combined (-81.4%)! The web is full of rare, repeated content that the LM can’t memorize in time. I solve this with retrieval-augmented n-gram hints🧵
90
238
61
2,888
300,171
Chao-Chun (Joe) Hsu retweeted
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵
539
1,258
738
9,048
2,401,188
Chao-Chun (Joe) Hsu retweeted
it’s my nature to keep posts rather impersonal, so i have to work extra hard to tweet this sort of personal testimony below from time to time as i think the reminder in it around AI and the “value of human intelligence” is worth it as of the end of Oct 2026, i have not (yet - you may insist) had the *slightest* moment of questioning the impact of AI on the meaning of any of my work/skills - or even feeling any unusual shift beyond having an amazing (and rather often infuriating) tool believe it or not, but my ability to produce a completed artifact* (paper, model, open source software, lecture, …) and the effort, expertise, or time one takes to produce has actually barely reduced - despite comprehensive productivity benefits from AI the reason is that i essentially NEVER choose to work for a pre-determined outcome or some field-wide well-specified problem; i only work on artifacts i aim to perfect according to *my* judgment and where my personalized ambition obviously has no set limit and isn’t somehow pre-loaded onto AI better tools simply change the costs of my choices - i quite literally work on moving goal-posts if you will! when the fixed constraint are only time and energy, AI simply allows me to re-explore scope and make fewer (or different!!) tradeoffs or compromises, which feels almost no different! i’m not saying every job or every endeavor - including very advanced and historically prized ones! - looks like this today, but i’m offering: (1) at least one voice in this moment (of so many annoying loud FUD voices!) that it’s totally possible and i suspect extremely scalable and i predict will remain so for the foreseeable future! (2) why “someone like me” is indeed choosing happily to be in academia and not elsewhere! (again lots of FUD out here!, but where else do you have the freedom to keep seeking to create value for deciding and solving ~arbitrary ambitious/useful problems, establish credit for it, and train superstars who do the same?) my fear and worry in this moment is a lot less about Al or "the value of human intelligence" and much more about how concentration of power has a very reliable history of being very dangerous to large numbers of affected people! a VERY different problem imo *i guess certain of those took way longer than i expected but do stay tuned for a half dozen things on the short horizon…
confession: i've yet to have that moment of feeling that frontier LLMs are geniuses, or super-human at ~anything broadly coherent that i deeply prize
28
38
7
532
45,790
Chao-Chun (Joe) Hsu retweeted
> add an SAE on top of a good retriever > the concepts want to shout > you can retrieve based on the concepts very exciting piece of work! I think interpretability will end up being a major contributor to the field of retrieval
We surfed the dense and late-interaction waves… now it's time to SlideOn sparse retrieval 🐕 🚀 Meet SlideOn and mSlideOn, fully open SOTA sparse retrievers at 174M and 339M parameters, trained by @LightOnIO R&D Models: huggingface.co/collections/l… Blog: huggingface.co/blog/lightona…
1
4
31
1,423
Chao-Chun (Joe) Hsu retweeted
We surfed the dense and late-interaction waves… now it's time to SlideOn sparse retrieval 🐕 🚀 Meet SlideOn and mSlideOn, fully open SOTA sparse retrievers at 174M and 339M parameters, trained by @LightOnIO R&D Models: huggingface.co/collections/l… Blog: huggingface.co/blog/lightona…
7
25
7
76
7,798
Chao-Chun (Joe) Hsu retweeted
I mean, when you know the authors you already know it's going to be insane, but honestly the blog post was so interesting Stellar work, and very timely: the future is sparse, and getting strong sparse models easily with training SAE on top of dense/LI models is the way
We surfed the dense and late-interaction waves… now it's time to SlideOn sparse retrieval 🐕 🚀 Meet SlideOn and mSlideOn, fully open SOTA sparse retrievers at 174M and 339M parameters, trained by @LightOnIO R&D Models: huggingface.co/collections/l… Blog: huggingface.co/blog/lightona…
3
4
38
2,010