Colendi AIResearch
Menu

Research archive

The Intelligence Papers

A navigable record of the ideas, artifacts, and institutions that formed the modern intelligence stack—reviewed as evidence, not arranged as a reading list.

281reviewed records

22research domains

25 records

0012023Agents & Tool Use

Peer reviewed

ReAct: Synergizing Reasoning and Acting in LMs

Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, Yuan Cao

The method prompts a language model to alternate between writing a reasoning step and issuing an action, such as a query to a search API, then feeding the observation back before the next thought. This lets the model plan, look things up to correct itself, and reduce fabricated facts on knowledge tasks like HotpotQA and FEVER, while the reasoning traces make its behavior on interactive benchmarks more legible. It became a common template for tool-using and agentic prompting because it needs no fine-tuning, only few-shot exemplars.

UnknownDifficulty 5/10Verified
Read record Primary source
0022023Agents & Tool Use

Peer reviewed

Toolformer: LMs Can Teach Themselves to Use Tools

Timo Schick, Jane Dwivedi-Yu, Roberto Dessi, et al.

The model samples candidate API calls (calculator, search, calendar, question-answering, translation) inside ordinary text, executes them, and keeps only the calls whose returned results lower the perplexity of the following tokens, producing a training set it then fine-tunes on. As a result a 6.7B-parameter model learns to invoke tools where they help and gains on arithmetic, factual QA, and temporal tasks without degrading its core language modeling. The mechanism showed tool use can be bootstrapped from the model's own predictions rather than curated examples.

UnknownDifficulty 5/10Verified
Read record Primary source
0032024Agents & Tool Use

Peer reviewed

SWE-agent / OpenHands (autonomous software engineering)

John Yang, Carlos E. Jimenez, Ofir Press, Karthik Narasimhan, Xingyao Wang, Graham Neubig

These systems give the model a defined set of actions for navigating files, editing code, and running commands, plus the feedback it gets back from executing those actions, so it can work through a real bug or feature request over many steps. On the SWE-bench benchmark of actual GitHub issues, this agent-plus-tools approach resolved a meaningful fraction of tasks that require reading a codebase, making changes, and checking them. The work showed that the design of the agent's action interface, not just the underlying model, strongly affects how many software tasks get solved, and the released open frameworks made this style of coding agent broadly available.

UnknownDifficulty 6/10Verified
Read record Primary source
0042023Agents & Tool Use

Peer reviewed

Generative Agents: Interactive Simulacra of Human Behavior

Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S. Bernstein

Generative agents combine large language models with a memory stream architecture that records experiences, retrieves relevant memories by recency/importance/relevance, reflects to form generalizations, and plans actions recursively. A society of 25 agents in a Sims-like sandbox successfully coordinated complex social behaviors like spreading party invitations and forming relationships, all from a single seed suggestion. This work demonstrated how memory, reflection, and planning enable agents to maintain coherent long-term behavior beyond a single context window.

Fully open sourceDifficulty 5/10Verified
Read record Primary source
0052023Agents & Tool Use

Product release

Function Calling / MRKL / Model Context Protocol

Ehud Karpas, Opher Lieber, Barak Lenz, Yoav Levine, Yoav Shoham, OpenAI, Anthropic

MRKL proposed a neuro-symbolic architecture in which a router directs a query to discrete expert modules, such as a calculator, database, or API, and merges their outputs. Later function-calling formats let a model emit a structured call that host code executes and feeds back, and the Model Context Protocol standardizes how tools, data sources, and context are exposed to models across applications. The common contribution is a defined boundary and calling convention between the model and outside systems, which enabled reliable access to current data, exact computation, and actions the model cannot perform internally.

UnknownDifficulty 4/10Verified
Read record Primary source
0062023Agents & Tool Use

Peer reviewed

Reflexion: Language Agents with Verbal Reinforcement Learning

Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, Shunyu Yao

After each trajectory, an evaluator produces a reward or error signal, and the agent generates a natural-language reflection on what went wrong, which is appended to an episodic memory buffer that conditions subsequent trials. Because learning happens through text in the context window rather than gradient updates, the same frozen model improves across attempts on decision-making, coding, and reasoning benchmarks. This provided a cheap iterative-improvement mechanism for agents that reuses a model's own language ability as the update rule.

UnknownDifficulty 5/10Verified
Read record Primary source
0072023Agents & Tool Use

Preprint

Voyager / Generative Agents

Guanzhi Wang, Linxi Fan, Anima Anandkumar, Joon Sung Park, Percy Liang, Michael S. Bernstein

Voyager runs a GPT-4 agent in Minecraft with an automatic curriculum that proposes progressively harder tasks, a growing library of executable code skills it writes and reuses, and an iterative prompting loop that repairs failures, letting it explore without a fixed objective. Generative Agents give each simulated character a memory stream of observations, a retrieval scheme, and periodic reflection that synthesizes higher-level inferences, producing emergent social behavior like spreading information and coordinating events. Together they showed that lifelong skill acquisition and persistent, plausible multi-agent behavior can be driven largely by language models plus structured memory rather than task-specific training.

UnknownDifficulty 5/10Verified
Read record Primary source
0082021Agents & Tool Use

Preprint

WebGPT: Browser-assisted Question-answering with Human Feedback

Reiichiro Nakano, Jacob Hilton, Suchir Balaji, John Schulman

WebGPT gives GPT-3 a text-based browser and trains it with human comparisons to search, navigate, and quote evidence, then optimizes against a learned reward model. The result answers long-form questions with references that raters prefer, at the cost of sometimes over-trusting sources. It was an early demonstration of tool use and retrieval-augmented answering with RLHF.

UnknownDifficulty 6/10Verified
Read record Primary source
0092022Agents & Tool Use

Preprint

Do As I Can, Not As I Say: Grounding Language in Robotic Affordances (SayCan)

Michael Ahn, Anthony Brohan, Noah Brown, Karol Hausman

SayCan scores candidate skills by multiplying an LLM’s estimate that a skill helps achieve the instruction with a value function’s estimate that the skill is currently feasible, so the robot picks useful and possible actions. This let a robot carry out long natural-language instructions in a kitchen. It was an influential early bridge from language models to embodied action.

UnknownDifficulty 6/10Verified
Read record Primary source
0102023Agents & Tool Use

Preprint

Agent Memory & Multi-Agent Systems (MemGPT/AutoGen/CAMEL)

Charles Packer, Joseph E. Gonzalez, Qingyun Wu, Chi Wang, Guohao Li, Bernard Ghanem

MemGPT borrows virtual-memory ideas, treating the context window as limited RAM and external storage as disk, so the agent pages information in and out under its own control to maintain long-running conversations and documents beyond the token limit. AutoGen provides a framework for defining multiple conversable agents that message each other and call tools to solve a task jointly, and CAMEL uses role-playing prompts to make two agents cooperate through dialogue with minimal human steering. The shared contribution is infrastructure for memory that outlives a single context and for coordinating several agents, enabling longer and more complex workflows than a single stateless call.

UnknownDifficulty 6/10Verified
Read record Primary source
0142025Agents & Tool Use

Preprint

SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution

Yuxiang Wei, Olivier Duchenne, Jade Copet, Quentin Carbonneaux, Lingming Zhang, Daniel Fried, Gabriele Synnaeve, Rishabh Singh, Sida Wang (Meta)

Uses rule-based rewards derived from GitHub commit histories to train issue-fixing coding agents by RL on real software evolution, without any proprietary model in the pipeline.

UnknownDifficulty 6/10Verified
Read record Primary source
0152025Agents & Tool Use

Preprint

Kimi K2: Open Agentic Intelligence

Kimi Team (Moonshot AI)

A trillion-parameter open Mixture-of-Experts model post-trained specifically for agentic tool orchestration, topping open-model agentic benchmarks (tau-bench, SWE-bench Verified) and making capable agent models openly available.

UnknownDifficulty 6/10Verified
Read record Primary source
0182025Agents & Tool Use

Technical report

Computer-Using Agent (CUA) / Operator

OpenAI

A GUI agent trained with reinforcement learning that combines vision and reasoning to operate websites through the interface rather than APIs, reaching state-of-the-art on OSWorld, WebArena, and WebVoyager.

UnknownDifficulty 6/10Verified
Read record Primary source
0192025Agents & Tool Use

Technical report

Introducing Deep Research

OpenAI

An agentic system built on a browsing-optimized o3 that autonomously runs multi-step web research and synthesizes analyst-grade reports, defining the deep-research agent category.

UnknownDifficulty 6/10Verified
Read record Primary source
0202025Agents & Tool Use

Technical report

How We Built Our Multi-Agent Research System

Jeremy Hadfield, Barry Zhang, Kenneth Lien, Florian Scholz, Jeremy Fox, Daniel Ford (Anthropic)

Documents a production orchestrator-worker architecture in which a lead agent spawns parallel subagents with separate context windows, plus the delegation and evaluation engineering that makes multi-agent research reliable.

UnknownDifficulty 6/10Verified
Read record Primary source
0222025Agents & Tool Use

Preprint

Recursive Language Models

Alex L. Zhang, Tim Kraska, Omar Khattab

Recursive Language Models treat the user prompt as a variable in a REPL environment rather than feeding it directly into the neural network. The LLM writes code to peek into and decompose the prompt, and iteratively invokes itself recursively over programmatic snippets. This paradigm processes inputs more than 10x beyond context limits while dramatically outperforming vanilla frontier LLMs and common long-context scaffolds, maintaining strong performance even at 10M+ token scale with comparable cost.

Fully open sourceDifficulty 7/10Verified
Read record Primary source
0232025Agents & Tool Use

Preprint

A-Mem: Agentic Memory for LLM Agents

Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, Yongfeng Zhang

A-Mem extends agent memory beyond simple storage and retrieval by letting agents dynamically create comprehensive notes with contextual descriptions, keywords, and tags, then autonomously establish connections based on semantic similarities. As new memories are integrated, they can trigger updates to existing memories, allowing the knowledge network to continuously refine itself. This approach enables more adaptive memory management that generalizes better across diverse tasks than fixed-structure memory systems.

Fully open sourceDifficulty 6/10Verified
Read record Primary source
0242025Agents & Tool Use

Preprint

Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory

Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, Deshraj Yadav

Mem0 addresses the fixed context window limitation by implementing extraction and update phases that identify key facts from conversations and evaluate them against existing memories, applying ADD/UPDATE/DELETE/NOOP operations. An enhanced variant with graph-based memory representations captures complex relational structures. Evaluated on the LOCOMO benchmark, Mem0 achieved 26% improvement over OpenAI while reducing p95 latency by 91% and saving over 90% in token costs compared to full-context approaches.

Fully open sourceDifficulty 5/10Verified
Read record Primary source
0252023Agents & Tool Use

Software release

AutoGPT / BabyAGI (viral agent software)

Toran Bruce Richards, Yohei Nakajima

Given a high-level goal, these programs prompt the model to generate a task list, execute tasks one at a time using tools like web search and file access, and feed results back to reprioritize and spawn new tasks, running with little human intervention. BabyAGI centered on a compact task-creation and prioritization loop backed by a vector store, while AutoGPT added tool integrations and persistence around a similar cycle. They were engineering demonstrations rather than research papers, and their viral spread popularized the autonomous-agent pattern while also exposing its practical limits in reliability, looping, and cost.

Source availableDifficulty 3/10Verified
Read record Primary source