Research Paper

Introduction Post-training is where the real cost of LLM development lives. Taking a pretrained base model and turning it into something actually useful - an assistant that follows instructions, reasons carefully, and behaves safely - requires months of supervised fine-tuning, reward modeling, and alignment work from teams of skilled ML

Last month Anthropic published something unusual: a detailed accusation. Three Chinese AI labs, DeepSeek, Moonshot AI, and MiniMax, had collectively generated over 16 million exchanges with Claude through ~24,000 fraudulent accounts. They bypassed regional access controls using commercial proxy networks. They targeted specific, high-value capabilities. Anthropic called it a

Introduction Every team building on top of LLMs has a version of the same mental model: pretraining teaches the model what it knows, and post-training teaches it how to behave. Don't want it to say harmful things? Train that out. Want it to be more helpful? Push that

Introduction If you follow the LLM space, you've probably heard a lot about parameter counts, context windows, and benchmark scores. What gets discussed far less often is the mechanism that makes all of it possible: attention. Every major language model (GPT, Llama, Gemini, Qwen, DeepSeek) is built on

There's a quiet assumption baked into most of the recent excitement around reasoning models: that the impressive gains you see from systems like DeepSeek-R1 or similar RL-trained models come from something genuinely new: novel capabilities that the base model simply didn't have before training. A new

Introduction Voice AI hit a genuine inflection point when full-duplex models arrived. Systems like Moshi finally cracked the core problem with conversational speech: the awkward cascade of listen, then transcribe, then think, then speak. Full-duplex models [models that listen and speak simultaneously over a continuous audio stream, the same way

I recently came across a fascinating paper at ICLR’26 that tackles a question many of us AI developers have been wrestling with: can we trust LLMs to evaluate complex, interactive task? The authors focus on the domain of web development, and the short answer: we've got a

Introduction Every mainstream large language model today generates text the same way: one token at a time, left to right, no looking back. It works remarkably well, but it has a structural flaw that's easy to overlook until you care about speed at scale. The model can never

When was the last time you confidently trusted a benchmark to tell you how an LLM would actually perform in production? The gap between benchmark performance and real-world reliability is significant, and it's a problem that deserves more attention. I recently read through this paper by Inference Labs,

Have you ever noticed how some voice assistants seem to understand you instantly, while others leave you waiting? That delay isn't random. It's the result of a fundamental trade-off that has plagued speech recognition for years. Systems could either be fast or accurate, but rarely both.

Recent research from the Zilliz team tackles a problem that shows up constantly in production RAG systems: how do you actually show users why a document is relevant to their query? Consider a typical scenario. A user asks: "How can I speed up my Python code?" The vector

Introduction Moonshot AI recently open-sourced Kimi K2 and its reasoning-optimized variant, K2 Thinking. As someone who works with large language models, I wanted to break down what makes this release interesting and where it pushes forward the state of open-source AI. K2 Thinking is a 1-trillion parameter model that can