Home›Papers
Papers
Papers
Papers that changed AI, with a plain-language explainer for each.
Total 21 · Clear filters
- 2026
Muse Spark Safety & Preparedness Report
arXiv
Meta's safety and preparedness report for Muse Spark, the model behind Meta AI, sets out the catastrophic-risk evaluations behind its launch and argues that, with mitigations applied, residual risk is acceptable.
- 2026
GLM-5: from Vibe Coding to Agentic Engineering
arXiv
Zhipu's GLM-5 report describes a 744-billion-parameter mixture-of-experts model built on sparse attention and asynchronous reinforcement learning, aimed at moving AI-assisted programming from vibe coding toward agentic engineering.
- 2026
Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
arXiv
Terminal-Bench 2.0 is a curated benchmark of 89 hard tasks carried out in computer terminal environments, built to measure how well AI agents handle realistic command-line work; frontier models and agents score below 65% on it.
- 2025
Kimi K2: Open Agentic Intelligence
arXiv
Kimi K2 is Moonshot AI's trillion-parameter mixture-of-experts model, pre-trained with the MuonClip optimizer and post-trained on large-scale agentic data plus reinforcement learning; the report reports state-of-the-art results among open-source non-thinking models and releases base and post-trained checkpoints.
- 2024
τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains
arXiv
τ-bench is a benchmark, submitted to arXiv in June 2024, that tests language agents in dynamic conversations with a simulated user, with domain-specific API tools and policy rules, and scores them by database state rather than by the text of their replies.
- 2024
OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
arXiv
OSWorld is a benchmark and environment for testing multimodal agents on 369 real computer tasks in Ubuntu, Windows and macOS; humans complete over 72.36% of the tasks, while the best model reaches only 12.24%.
- 2024
LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
arXiv
A benchmark for code LLMs that keeps collecting fresh contest problems from LeetCode, AtCoder and CodeForces, so that scores reflect ability rather than memorised test data, and that also tests code repair and execution reasoning.
- 2023
GAIA: a benchmark for General AI Assistants
arXiv
GAIA is a benchmark of real-world questions that are easy for humans but hard for AI assistants, testing reasoning, multimodality, web browsing, and tool use. It compares AI progress against average human performance.
- 2023
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
arXiv
A benchmark of 2,294 real software engineering problems taken from GitHub issues and the pull requests that resolved them, in which a model must edit a codebase to fix an issue; on release the best model solved under 2% of them.
- 2023
WebArena: A Realistic Web Environment for Building Autonomous Agents
arXiv
WebArena is a reproducible environment of four fully functional websites — e-commerce, a social forum, collaborative software development and content management — built for testing language-guided web agents. Its best GPT-4 agent completed 14.41% of tasks; humans completed 78.24%.[^1]
- 2023
Toolformer: Language Models Can Teach Themselves to Use Tools
arXiv
A 2023 paper by Schick, Dwivedi-Yu, Dessì and co-authors showing that a language model can teach itself when to call a calculator, search engine, Q&A system, translation system or calendar, and how to use the results.
- 2022
Constitutional AI: Harmlessness from AI Feedback
arXiv
Constitutional AI, a 2022 paper from Anthropic, proposes training a harmless AI assistant with AI-generated feedback guided by a written set of principles, instead of human labels of harmful outputs.
- 2022
Holistic Evaluation of Language Models (HELM)
arXiv
A 2022 paper by Percy Liang, Rishi Bommasani, Tony Lee and many co-authors that ran 30 language models through the same 42 scenarios and seven metrics, turning scattered benchmarks into one standardized evaluation.
- 2022
ReAct: Synergizing Reasoning and Acting in Language Models
arXiv
ReAct is a prompting method that has a language model interleave written reasoning with calls to outside tools or environments, improving question answering, fact verification and interactive decision-making.
- 2022
Training language models to follow instructions with human feedback (InstructGPT)
NeurIPS
InstructGPT used human-written demonstrations, human rankings of outputs and reinforcement learning to make GPT-3 follow user instructions better. Labelers preferred its outputs even though it was far smaller than GPT-3.
- 2021
Program Synthesis with Large Language Models (MBPP)
arXiv
A 2021 arXiv paper by Austin, Odena, Nye and eight other authors that introduced MBPP, a benchmark of 974 short Python programming tasks, and measured how language models from 244M to 137B parameters turn natural-language descriptions into working code.
- 2021
Evaluating Large Language Models Trained on Code (Codex / HumanEval)
arXiv
A 2021 paper from OpenAI introduces Codex, a GPT model fine-tuned on public GitHub code, and HumanEval, an evaluation set that judges generated Python programs by whether they work rather than by how closely they match a reference.
- 2021
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
FAccT '21
A 2021 FAccT paper by Bender, Gebru, McMillan-Major and Shmitchell that asks whether language models can be too large and sets out recommendations on costs, data, research goals and alternative directions.
- 2020
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (RAG)
arXiv
A 2020 paper that pairs a pre-trained seq2seq model with a dense Wikipedia index fetched by a neural retriever, setting state of the art on three open-domain QA tasks and producing more factual text than a parametric-only baseline.
- 2019
Language Models are Unsupervised Multitask Learners (GPT-2)
OpenAI
OpenAI's 2019 paper on GPT-2, a 1.5-billion-parameter Transformer trained on web text. It showed that a model trained only to predict the next word could perform several tasks when prompted, and its staged release sparked public debate.