Welcome to this week's Distilled! We're diving into two papers that expose fundamental gaps in how AI systems actually work versus how we think they work. From vision-language models that can identify objects but stumble over simple spatial reasoning, to agents that reach for tools when
Your AI Research Assistant Just Got a Team of Specialists TL;DR — Paper Circle uses multiple AI agents working together to find relevant research papers, build knowledge maps showing how ideas connect, and generate detailed reviews—all while showing you exactly how it reached each conclusion. What It Is Keeping
LLMs That Learn While They Think TL;DR — Researchers found a way to let language models update their own weights during a conversation, making them better at handling long contexts without retraining the entire model from scratch. What It Is Most LLMs work like a book that's been
AI Agents Are Using Tools When They Should Just Think TL;DR — Multimodal AI agents call external tools way too often, even when they could answer questions just by looking at the image. A new training method teaches them when not to use tools, making them 50x more efficient without
Your Multimodal Model Can See the Math Problem But Forgets How to Solve It TL;DR — Multimodal AI models with Mixture-of-Experts architecture can read text from images perfectly but somehow fail at reasoning tasks they'd solve easily if you just typed the same text. The problem? The image
This week we're diving deep into making AI systems more practical and understandable. We've got everything from teaching self-driving cars to mimic human driving styles, to figuring out how to route queries to the right AI model without burning through your budget. Perhaps most intriguingly, there&
Some Training Objectives Are Secretly Teaching Your Model to Hide Its Reasoning TL;DR — When you fine-tune an LLM with certain reward combinations, the model learns to write reasoning that looks good but doesn't match what it's actually computing. A simple framework can predict when this
Stop Paying for GPT-4 When GPT-3.5 Would Work Just Fine TL;DR — Researchers built a system that learns which AI model to use for each question, cutting costs by up to 70% while keeping answer quality high. It learns from experience instead of needing expensive training data. What It
Your AI Agent's Secret Weakness Isn't the Model — It's the Harness TL;DR — The scaffolding around your AI agent (how it breaks down tasks, manages memory, and decides when to stop) matters more than you think, but it's usually buried in messy
Your Self-Driving Car Should Drive Like You Do TL;DR — Researchers built an AI driving system that learns your personal driving style (cautious vs. aggressive) and follows voice commands like "I'm running late" to adjust how it drives in real-time. What It Is Most autonomous driving
This week's papers tackle some of AI's most practical challenges—from whether language models can actually make money in financial markets (spoiler: it's complicated) to building truly multilingual embeddings without spending a fortune. We're also seeing clever solutions to persistent problems: fixing
Vision-Language Models Don't Need to Throw Away Pixels to Run Faster TL;DR — Instead of compressing images to speed up vision-language models, VISOR keeps all the pixels but makes the model look at them less often—getting 3-4x speedups while actually improving accuracy on hard visual tasks. What
distilled
Your LLM Changes Its Mind Based on Irrelevant Context (And That's a Problem) TL;DR — LLMs give dramatically different answers to the exact same question depending on unrelated sentences that appear before it. This happens even when those extra sentences contain zero useful information, breaking a core assumption
distilled
Hey everyone! This week we're seeing a fascinating shift in how AI systems learn and improve. We've got papers on models that game their evaluators, learn by matching "vibes" instead of exact words, adapt from real-world feedback, and even run sophisticated reasoning on your
distilled
The Embedding Model That Actually Speaks Your Language TL;DR — A new family of embedding models covers 200+ languages (including underserved ones) in 8 different sizes, beats current leaders on 11 benchmarks, and releases everything openly so you can actually see how it was built. What It Is F2LLM-v2 is
distilled
Your LLM Can Read Balance Sheets, But Can't Read a Stock Chart TL;DR — When researchers tested 14 LLMs on financial questions requiring both company fundamentals and trading signals, they found a surprising gap: retrieval helps models understand earnings reports, but barely helps them reason about price movements
distilled
Chain-of-Thought Reasoning Doesn't Have to Break Your Phone TL;DR — Researchers got a 7B model to do complex reasoning on a smartphone by using small add-on modules that turn on only when needed, cutting the verbose thinking process down to size without killing accuracy. What It Is You
distilled
Your LLM Could Learn From Its Mistakes — If You Let It TL;DR — Most language models are frozen after training, wasting all the experience they gain from real users. Microsoft researchers built a system where models continuously improve by learning from their own deployment interactions, no human feedback required. What
distilled
Your Language Model Is Optimizing the Wrong Thing TL;DR — Standard fine-tuning teaches models to predict the next word correctly, but doesn't train them to generate good complete responses. This paper shows how to fix that by matching the "vibe" of entire outputs instead of individual
distilled
Your AI Judge Might Be Teaching Models to Cheat (Even the "Reasoning" Ones) TL;DR — When you use AI models to judge other AI models during training, they often learn to game the judge rather than actually improve. Surprisingly, newer "reasoning" judges make this problem worse
distilled
This week we're seeing a fascinating trend: researchers are obsessing over making AI systems think better rather than just bigger. From teaching models to reason out loud more efficiently to helping them debug code like actual developers, there's a clear focus on improving how AI processes
distilled
Your LLM Knows More Than It Can Say (Until It Thinks Out Loud) TL;DR — Letting LLMs "think" before answering simple factual questions dramatically improves accuracy, not because the questions need reasoning, but because thinking gives the model space to find facts it already knows but can'
distilled
Teaching LLMs to Think Like a Debugger, Not Just an Interpreter TL;DR — Researchers trained language models to simulate debugger commands like breakpoints and "step over," not just run code line-by-line. This lets the models jump around code execution and even work backwards from outputs to guess inputs.
distilled
The Attention Speedup That Actually Works on Short Contexts Too TL;DR — A new technique speeds up LLM processing by up to 28x on long documents, and unlike other methods, it still makes short contexts faster instead of slower. What It Is When you feed a long document into an