Papers22

SLATE is a new way to teach AI to think step by step while using a search engine, giving feedback at each step instead of only at the end.

Tool-R0 teaches a language model to use software tools (like APIs) with zero human-made training data.

This paper teaches a language model to improve its own math answers by first writing several drafts and then learning to beat its best draft.

Most reinforcement learning agents only get a simple pass/fail reward, which hides how good or bad their attempts really were.

This paper shows that many reasoning failures in AI are caused by just a few distracting words in the prompt, not because the problems are too hard.

This paper introduces Foundation-Sec-8B-Reasoning, a small (8 billion parameter) AI model that is trained to “think out loud” before answering cybersecurity questions.

LLM agents are usually trained in a few worlds but asked to work in many different, unseen worlds, which often hurts their performance.

Typhoon-S is a simple, open recipe that turns a basic language model into a helpful assistant and then teaches it important local skills, all on small budgets.

Academic rebuttals are not just about being polite; they are about smart, strategic persuasion under hidden information.

Small AI models often stumble when a tool call fails and then get stuck repeating bad calls instead of fixing the mistake.

Diffusion language models can write tokens in any order, but that freedom can accidentally hurt their ability to reason well.

Think3D lets AI models stop guessing from flat pictures and start exploring real 3D space, like walking around a room in a video game.