Reviewers automatically assume that simple is not novel. This is sheer laziness. Yes, it may be simple and obvious in retrospect, but someone had to have that insight first. Simple is good. Simple is robust, easy to implement and reproduce, broadly applicable, etc.
Jimmy Lin
4,472 posts
I profess CS-ly at @UWaterloo. Previously, I monkeyed code for @Twitter, slides for @Cloudera, and scienced for @yupp_ai.
- Itโs been 36 hours since Grok 4 launched and we have an early verdict based on 6K+ preferences of @yupp_ai users globally on real use cases. โผ๏ธ Grok 4 is worse than other leading models: OpenAI o3, Claude Opus 4, and Gemini 2.5 Pro. Grok 4 is liked even less than Grok 3. ๐งต
- DAAM... You saw it here first! Attribution maps for Stable Diffusion based on upscaling and aggregating cross-attention activations in the latent denoising subnetwork: arxiv.org/abs/2210.04885 example for "an angry, bald man doing research" below - demo at daam.ralphtang.com:8080
- GPT-4 and its ilk are awesome for rapid prototyping and one-offs, but at the end of the day, enterprises will deploy far smaller distilled models in production. Here's my contrarian take -
- So, CV researchers are looking at transformers and NLP researchers are looking at CNNs (again). What a strange world.
- Still cropping and modifying BERT diagrams from Devlin et al. (2019)? I spent several hours redrawing BERT in PowerPoint so you don't have to... Perfect for use in presentations, papers, etc.! Happy to share under Releasing under CC BY 4.0 cs.uwaterloo.ca/~jimmylin/BERTโฆ
- Following the AI Residency program by Google, Facebook, Microsoft, Uber, etc., I'd like to start the Waterloo AI Residency program. It's called grad school.
- "NLP makes IR interesting and IR makes NLP useful!" - slides from my #sigir2020 summer school talk at: cs.uwaterloo.ca/~jimmylin/publโฆ Get your rotten tomatoes and eggs out!
- BERT is three years old today!
- Happy to share an early draft of "Pretrained Transformers for Text Ranking: BERT and Beyond", our forthcoming book (tentatively, early 2021) by @lintool @rodrigfnogueira @andrewyates
- For vector search, practitioners kinda know that for small corpora, don't bother with HNSW indexing, just brute-force it. However, guidance is mostly hand wavy... until now. I ran some experiments for you on BEIR and wrote it up. arxiv.org/abs/2409.06464 You're welcome.
- My (contrarian?) take: prompt engineering is programming in natural language. We've tried this before, with attempts dating back decades. Recent advances do not change the fact that natural languages are ambiguous, imprecise, under-specified, highly contextual, etc.
- Recently, @CohereAI boasted "3X better performance" in multilingual text understanding. We tested that claim by evaluating Cohere embeddings on MIRACL: tl;dr - We weren't able to replicate the 3X claim, but we did observe a 38% improvement over BM25. github.com/castorini/pyseโฆ





