Breaks hard reasoning problems into simpler subproblems, then solves them using earlier answers. With 14 exemplars, code-davinci-002 reaches at least 99% accuracy on SCAN splits, including the length split.
AI Dataset2022
Contains 40,000 teacher-generated reasoning traces distilled from the Qwen3.8-27B model for supervised fine-tuning and analysis. Covers code, math, science and logic; each example pairs a <think> chain-of-thought with a final response and is distributed in JSONL/Parquet for SFT workflows.
Splits complex tasks into reusable subtasks, each with its own prompt or symbolic handler. Recursive decomposition and retrieval improve GPT-3 reasoning on symbolic tasks and multi-hop question answering.
AI Dataset2023
Benchmark for evaluating general AI assistants with 466 short, real-world questions that require tool use, multimodality and reasoning; provides a public dev set and a withheld test set used for leaderboard evaluation.
Improves language-model problem solving by generating candidate answers, retaining verified successes, and fine-tuning on them. Repeats this loop with an expectation-maximization interpretation instead of collecting more human-written solutions.
AI Dataset2024
A 12K-question multiple-choice benchmark that raises difficulty by expanding choices to ten and adding reasoning-focused problems. Designed for robust LLM evaluation with cross-domain coverage, expert curation, and emphasis on chain-of-thought performance.
A concise textbook-style book that explains foundational concepts and techniques for large language models, covering pre-training, generative models, prompting, alignment, inference, and reasoning. Structured as self-contained chapters for readers with some ML/NLP background or those seeking a principled introduction to LLM foundations.
AI Agent2025
Autonomously executes diverse biomedical research tasks by combining LLM reasoning, retrieval-augmented planning, and code-based execution. Includes a web UI and Gradio demo, a curated Know‑How library, MCP integration, and a biology-tailored reasoning model (Biomni‑R0).
Explains how base-model behavior and GRPO normalization distort claims about reasoning gains. Introduces Dr. GRPO to reduce length bias, with a 7B training recipe reaching 43.3% on AIME 2024.
AI Agent Papers2025
Trains web agents to gather information through longer interaction, using an online RL curriculum that expands rollout horizons. Tests on WebVoyager and WebArena show how exploration can complement longer reasoning before each action.
AI Infra2025
Provides a deterministic context and knowledge-graph layer under LLMs and vector stores to record auditable decisions, provenance, and explainable rule-based reasoning. Supports polyglot graph storage, W3C PROV-O export, SHACL governance, and self-hosted enterprise connectors.
AI Agent Papers2025
Turns human desktop demonstrations into training data for agents that reason over screenshots and control mouse and keyboard. Releases a cross-OS collection tool, AgentNet data, and models; the paper reports 45.0% on OSWorld-Verified for its 72B model.