LLM Agents Papers
A topic-organized reading list for research on large language model agents.
- Source repo
- zjunlp/LLMAgentPapers
- Stars
- ★ 3.1k
- Last updated
- 1d ago
- FA score
- Insufficient evidence
At a glance
- How it runs
- Works with
- Portable with changes
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- Typical use
- A researcher entering the LLM-agent field can build a reading path through surveys, memory, planning, and tool-use sections.
- Not a fit if
- Teams needing a runnable agent, API, or CLI
- Researchers needing systematic search and completeness guarantees
- Users needing a clear license for compliance
- Source review
- Insufficient evidence 8 safety controls not found
What does this agent do, and when should you use it?
LLM Agents Papers is a curated index of research papers and resources rather than an installable or executable agent. It organizes the literature into overviews, individual agents, multiple agents, applications, frameworks, and other topics. The individual-agent section further covers personality, memory, planning, tool use, and reinforcement-learning training, while the multi-agent section distinguishes collaborative, adversarial, and open-ended interactions. Its resources include benchmarks such as Mind2Web, AgentBench, TravelPlanner, OSWorld, and AppWorld. Entries generally provide titles, authors, dates, and paper links, with code links for some projects; the repository's delivery boundary is a static research catalog, not a model runtime, API, CLI, or hosted service.
The repository maintains a manually organized directory in its README. Readers navigate from Content into Overview, Agent, Multiple Agents, Application, Framework, Others, and Resources, then browse subtopics such as Personality, Memory, Planning, Tool use, RL training, Collaborative Exchanges, Adversarial Interactions, Casual/Open Conversations, and Benchmarks. Individual records link to an abstract page or PDF and identify the paper's authors and date; selected records also include a code link. The repository does not itself search papers, invoke models, plan actions, run evaluations, process datasets, or expose documented APIs, classes, commands, or generated outputs.
- A researcher entering the LLM-agent field can build a reading path through surveys, memory, planning, and tool-use sections.
- A student developing a multi-agent research topic can separately survey collaborative communication, debate-style interaction, and social simulation.
- An engineer selecting evaluation material can locate benchmarks such as AgentBench, OSWorld, TravelPlanner, and AppWorld.
- An author preparing a literature review can use the categorized titles, author lists, dates, and links to assemble a candidate bibliography.
- A team surveying agent applications or frameworks can browse the dedicated sections and follow the code links included with selected papers.
How do you install or deploy this agent?
No installation is documented or required for the material shown. The source describes a GitHub-hosted paper list and provides no package, dependency manifest, build command, container image, service endpoint, or runtime requirement. The repository could ordinarily be cloned for offline access, but the supplied material does not document or verify a specific command as a product installation procedure.
How do you use this agent?
Open the repository URL, use the README Content index to select a research area, and follow each abs, paper, or code link to the underlying work. No credentials, configuration file, or startup command are documented for initial use. Automated search, filtering, export, citation management, and synchronization interfaces are not described; adopters needing those functions would have to build an integration or curate the records manually.
What are this agent's strengths and limitations?
- The taxonomy spans personality, memory, planning, tool use, reinforcement-learning training, and several forms of multi-agent interaction.
- Entries commonly combine a title, author list, paper link, and date, making it practical to identify and inspect primary papers.
- Applications, frameworks, and benchmarks receive dedicated sections, and selected papers include links to code.
- The supplied list ranges from early foundational work to entries dated 2026, giving readers a view across multiple stages of the field.
- This is a static reading list, not an operational agent; it provides no search API, CLI, model invocation, or automated workflow.
- The source does not document selection criteria, a systematic search method, peer-review verification, or completeness guarantees.
- The README contains duplicate numbering, skipped numbers, and inconsistent date formatting, so citations require manual verification.
- Repository metadata reports the license as unknown while the README displays an MIT badge; the supplied source does not include the license text needed to resolve that discrepancy.
- There is no documented installation, export, citation-management, or integration path for ChatGPT, Codex, Claude, or their APIs.
How does this agent compare with similar options?
The repository explicitly points to two adjacent lists from the same organization: Prompt4ReasoningPapers focuses on reasoning through language-model prompting, while KnowledgeEditingPapers focuses on knowledge editing for large language models. LLM Agents Papers instead centers on agent capabilities, multi-agent interaction, applications, frameworks, and evaluation resources.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| LLM Agents Papers This agent | Insufficient evidence | Library / SDKFree | ★ 3.1k | 1d ago | — | — |
| AI Research Skills | 77 · Some gaps | Agent plugin / skillFree + model costs | ★ 604 | 1d ago | Python | Claude Code |
| Zotero Research Assistant | 54 · Major gaps | Agent plugin / skillFree + model costs | ★ 3.5k | 3d ago | TypeScript | Codex · OpenAI API · Claude API |
| Rakazo AI Teammates | 63 · Some gaps | Self-hosted serviceFree + model costs | ★ 7.2k | today | TypeScript | — |
How does FollowAgents rate this agent?
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- On the supplied evidence, this repository is a bibliography rather than an installable or executable Agent product or framework; this assessment must not be treated as a safety or quality review of the listed papers or systems.
- The MIT badge alone does not establish the applicable license terms; reuse and redistribution rights should remain unconfirmed until the license text is available.
- Paper links, authors, dates, and classifications are recorded only from the supplied static text and were not executed, externally checked, or independently verified.