@lintool

I profess CS-ly at @UWaterloo. Previously, I monkeyed code for @Twitter, slides for @Cloudera, and scienced for @yupp_ai.

 Nearby data lake
cs.uwaterloo.ca/~jimmylin/
Joined February 2010
  • Tweets4.6K
  • Following875
  • Followers15.4K
438Media
tl;dr - the team @Sahel_Sharify @Lingwei_Gu @nour_jedidi speedran nanochat by @karpathy using ClimbMix by @shizhediao et al. and then SFT'ed on relevance data from @beirmug @crystina_z to arrive at a competitive pointwise reranker: arxiv.org/abs/2610.11922

Project Greenhouse: Progress Toward Fully Open and Sovereign Agentic Search

Project Greenhouse represents our exploration of a simple thesis: We believe that it is possible to build fully open and sovereign models for agentic search with only modest computational...

arxiv.org
3
10
41
3,333
It's about as good as SFT'ing Qwen, Gemma, etc. and just a bit worse than listwise GPT-6.1 Sol... all with just a handful of GPUs.
1
1
2
588
Our backbone: huggingface.co/castorini/gag… Our reranker: huggingface.co/castorini/gag…

castorini/gaggle-nanochat-pretrained-climbmix-20260924 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
1
1
496
What are we up to? 🏡 Project Greenhouse: Progress Toward Fully Open and Sovereign Agentic Search w/ @Sahel_Sharify @Lingwei_Gu @nour_jedidi arxiv.org/abs/2610.11922

Project Greenhouse: Progress Toward Fully Open and Sovereign Agentic Search

Project Greenhouse represents our exploration of a simple thesis: We believe that it is possible to build fully open and sovereign models for agentic search with only modest computational...

arxiv.org
1
19
43
2,955
Milestone 1️⃣ - we built a competitive pointwise decoder-only reranker completely from scratch - pre-training from ClimbMix + supervised fine-tuning on RLHN relevance data... with just a few GPUs... and it wasn't even that hard!
1
1
5
669
Fully open = transparent, sovereign = you're in control - You can do it too, let us show you how! Rigorous understanding FTW!
1
6
522
You mean, building things that are actually useful... and that people care about? Nice but how unacademic! 👍 Congratulations! I look forward to learning more! ❤️
Hamed Zamani
@HamedZamaniOct 7
We are releasing the most significant technology transfer effort by the CIIR in the last decade. Please share with your network. In this thread you will learn what Nage enables you to do. 🧵👇
1
13
1,720
Jimmy Lin retweeted
Dongfu Jiang
@DongfuJiangOct 5
PhD defended today! 🎓 Huge thanks to my supervisor @WenhuChen for all the guidance and support throughout this journey, and to all my committee members for their time, thoughtful questions, and feedback. Next stop: joining Meta FAIR as a Research Scientist. Excited for what’s ahead!
🤖 Made with AI
44
11
2
245
31,309
Jimmy Lin retweeted
Nour Jedidi@nour_jedidiOct 1
Excited to share that our paper was accepted to EMNLP Findings 🎉 We ask a simple question: Does reasoning improve pointwise rerankers? The answer? It can actually hurt... even for Rank1, a reranker trained to reason-then-answer 😱 📄 nourj98.github.io/papers/don…
Yung-Sung Chuang@YungSungChuang27 May 2025
🚨Do passage rerankers really need explicit reasoning?🤔—Maybe Not! Our findings: ⚖️Standard rerankers outperform those w/ step-by-step reasoning! 🚫Disable reasoning from reasoning reranker actually improves reranking accuracy!🤯 👇But, why? 📰arxiv.org/abs/2505.16886 (1/6)
2
8
25
1,973
Jimmy Lin retweeted
Nour Jedidi@nour_jedidiSep 30
To make an LLM a strong dense retriever, you often need training… But what if you only have a few-examples 🥩🥦 and no training budget? Just add RICE 🍚, a simple way to turn your LLM into an effective dense retriever that requires zero training 👨‍🍳 📄 arxiv.org/abs/2609.38099

Effective Dense Retrieval using Only In-Context Examples

Turning decoder-only large language models (LLMs) into strong dense retrievers typically requires some form of retriever training. In this paper, we ask whether LLMs can instead be prompted to...

arxiv.org
3
7
1
31
2,202
Yo, I heard you like leaderboards. So I created another one: with ACM Fellows and Turing Award winners. Another target for citation and h-index benchmaxxing. 💪 lintool.github.io/acm-bigcow…
1
13
4,549
I'm in a good mood because "grep is all you need" and "BM25 is all you need" both recently drew from (separate) Bernoullis and came back positive. Now I can claim that both have been +1ed by LLMs (wait, I meant peers). #nerdref
5
59
9,226
Summer passed in a blur... 🏖️🦩🍹 and I forgot to congratulate @HosnaOH for finishing her Master's! Her work with @amirhkarimi_ "A Unified Perturbation Framework for Analyzing Leaderboard Stability and Manipulation" arxiv.org/abs/2605.15761 just got a +1 from NeurIPS, so yay! 🎉
1
2
2
51
3,782
Achievement unlocked! 🔓 Does this come with stat bonuses or help with saving throws? uwaterloo.ca/computer-scienc…

Jimmy Lin appointed Fellow of the Royal Society of Canada | Cheriton School of Computer Science |...

An article detailing the achievements that led to Professor Jimmy Lin being named a 2026 Fellow of the Royal Society of Canada

uwaterloo.ca
36
9
1
147
6,656
My colleague @yuntiandeng about “Claudish” 🤖 and how AI coding agents are changing the way we communicate! uwaterloo.ca/news/media/q-an…

Q and A with the Experts: What is “Claudish”? | Waterloo News | University of Waterloo

University of Waterloo professor Yuntian Deng answers some questions about Claudish and his new translator.

uwaterloo.ca
1
1
1
10
2,721
1/ 🧵 BrowseComp-Plus (ACL 2026) set the standard for disentangled agentic search eval... 🙌 𝐁𝐫𝐨𝐰𝐬𝐞𝐂𝐨𝐦𝐩-𝐏𝐥𝐮𝐬_𝐂𝐌 makes it even better! 🚀 Same BCP questions grounded on 𝐂𝐥𝐢𝐦𝐛𝐌𝐢𝐱-𝟒𝟎𝟎𝐁 with 𝟓𝟓𝟑𝐌 𝐝𝐨𝐜𝐬. 💻 github.com/castorini/cmass

GitHub - castorini/cmass

Contribute to castorini/cmass development by creating an account on GitHub.

github.com
Xueguang Ma
@xueguang_ma11 Aug 2025
🚀 Introducing BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent. It is a new Deep-Research evaluation benchmark built on top of BrowseComp. It features - 📚 a fixed, carefully curated corpus of web documents - ✅ human-verified positive documents - ⚔️ web-mined challenging hard negatives. With BrowseComp-Plus, you can thoroughly evaluate and compare the performance of different components in a deep-research system. e.g. GPT-5 + Qwen3-Embedding. Code, dataset, and leaderboard links are provided at the end of this thread.
1
9
34
4,524
Jimmy Lin retweeted
Nandan Thakur@beirmugAug 11
In the past year, we've seen rapid progress in search agents! However, how well do these search agents work outside English? Introducing⛵️MAST @🔥FIRE 2026, a new shared task focused on multilingual agentic search.
1
8
3
31
6,341
Agentic reproducibility is the unlock, I claim. To what? 🙋‍♂️ Find out by reading my piece "Reproducibility in the Era of AI Agents" 🤖 in the June 2026 issue of SIGIR Forum. sigir.org/forum/issues/june-…
Jimmy Lin
@lintoolMay 11
What I'm cooking up... 👨‍🍳
1
6
1
30
3,897
Side note: I wrote this piece back in May, ~3 months ago. It's not aged too badly, I think? 🥛 Such is the speed of academic publishing...
645
Jimmy Lin retweeted
TREC RAG @ 2026@TREC_RAGAug 10
And with that, all TREC RAG 2026 submissions are in 🎉 We had an incredible turnout this year, with 30 teams submitting more than 80 RAG runs and 110 retrieval runs! A huge thank you to everyone who participated! We can’t wait to start evaluating your submissions 🚀
5
3
11
4,040
Load more