Difference between revisions of "AI Agents"

From GISAXS
Jump to: navigation, search
(→‎Related work)
(→‎Metrics, Benchmarks)
 
(2 intermediate revisions by the same user not shown)
Line 116: Line 116:
 
* 2024-12: [https://arxiv.org/abs/2412.05579 LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods]
 
* 2024-12: [https://arxiv.org/abs/2412.05579 LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods]
 
* 2025-03: [https://arxiv.org/abs/2503.19877 Scaling Evaluation-time Compute with Reasoning Models as Process Evaluators]
 
* 2025-03: [https://arxiv.org/abs/2503.19877 Scaling Evaluation-time Compute with Reasoning Models as Process Evaluators]
 +
* 2025-04: [https://arxiv.org/abs/2504.00050 JudgeLRM: Large Reasoning Models as a Judge]
  
 
===Deep Research===
 
===Deep Research===
Line 358: Line 359:
  
 
===Metrics, Benchmarks===
 
===Metrics, Benchmarks===
 +
See also: [[AI benchmarks]]
 
* 2019-11: [https://arxiv.org/abs/1911.01547 On the Measure of Intelligence]
 
* 2019-11: [https://arxiv.org/abs/1911.01547 On the Measure of Intelligence]
 
* 2022-06: [https://arxiv.org/abs/2206.10498 PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change]
 
* 2022-06: [https://arxiv.org/abs/2206.10498 PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change]
Line 381: Line 383:
 
* 2025-02: [https://arxiv.org/abs/2502.18356 WebGames: Challenging General-Purpose Web-Browsing AI Agents]
 
* 2025-02: [https://arxiv.org/abs/2502.18356 WebGames: Challenging General-Purpose Web-Browsing AI Agents]
 
* 2025-03: ColBench: [https://arxiv.org/abs/2503.15478 SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks]
 
* 2025-03: ColBench: [https://arxiv.org/abs/2503.15478 SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks]
 +
* 2025-04 OpenAI [https://openai.com/index/browsecomp/ BrowseComp: a benchmark for browsing agents]
  
 
===Evaluation Schemes===
 
===Evaluation Schemes===

Latest revision as of 12:00, 13 April 2025

Reviews & Perspectives

Published

Continually updating

Analysis/Opinions

Guides

AI Assistants

Components of AI Assistants

Agent Internal Workflow Management

Information Retrieval (Memory)

Contextual Memory

  • Memobase: user profile-based memory (long-term user memory for genAI) applications)

Control (tool-use, computer use, etc.)

Model Context Protocol (MCP)

Agent2Agent Protocol (A2A)

Open-source

Personalities/Personas

Specific Uses for AI Assistants

Computer Use

Software Engineering

Science Agents

See Science Agents.

Medicine

LLM-as-judge

Deep Research

Advanced Workflows

Streamline Administrative Tasks

Author Research Articles

Software Development Workflows

Several paradigms of AI-assisted coding have arisen:

  1. Manual, human driven
  2. AI-aided through chat/dialogue, where the human asks for code and then copies it into the project
    1. OpenAI ChatGPT
    2. Anthropic Claude
  3. API calls to an LLM, which generates code and inserts the file into the project
  4. LLM-integration into the IDE
    1. Copilot
    2. Qodo (Codium) & AlphaCodium (preprint, code)
    3. Cursor
    4. Codeium Windsurf (with "Cascade" AI Agent)
    5. ByteDance Trae AI
    6. Tabnine
    7. Traycer
    8. IDX: free
    9. Aide: open-source AI-native code editor (fork of VS Code)
    10. continue.dev: open-source code assistant
    11. Pear AI: open-source code editor
    12. Haystack Editor: canvas UI
    13. Onlook: for designers
    14. All Hands AI
    15. Devin 2.0 (Cognition AI)
    16. Google Firebase Studio
  5. AI-assisted IDE, where the AI generates and manages the dev environment
    1. Replit
    2. Aider (code): Pair programming on commandline
    3. Pythagora
    4. StackBlitz bolt.new
    5. Cline (formerly Claude Dev)
  6. Prompt-to-product
    1. Github Spark (demo video)
    2. Create.xyz: text-to-app, replicate product from link
    3. a0.dev: generate mobil apps (from your phone)
    4. Softgen: web app developer
    5. wrapifai: build form-based apps
    6. Lovable: web app (from text, screenshot, etc.)
    7. Vercel v0
    8. MarsX (John Rush): SaaS builder
    9. Webdraw: turn sketches into web apps
    10. Tempo Labs: build React apps
    11. Databutton: no-code software development
    12. base44: no-code dashboard apps
    13. Origin AI
  7. Semi-autonomous software engineer agents
    1. Devin (Cognition AI)
    2. Amazon Q (and CodeWhisperer)
    3. Honeycomb
    4. Claude Code

For a review of the current state of software-engineering agentic approaches, see:

Corporate AI Agent Ventures

Mundane Workflows and Capabilities

Inference-compute Reasoning

AI Assistant

Agentic Systems

Increasing AI Agent Intelligence

See: Increasing AI Intelligence

Multi-agent orchestration

Research

Organization Schemes

Societies and Communities of AI agents

Domain-specific

Research demos

Related work

Inter-agent communications

Architectures

Open Source Frameworks

Open Source Systems

Commercial Automation Frameworks

Multi-agent Handoff/Collaboration

Spreadsheet

Cloud solutions

Frameworks

Optimization

Reviews

Metrics, Benchmarks

See also: AI benchmarks

Evaluation Schemes

Multi-agent

Agent Challenges

  • Aidan-Bench: Test creativity by having a particular LLM generate long sequence of outputs (meant to be different), and measuring how long it can go before duplications appear.
  • Pictionary: LLM suggests prompt, multiple LLMs generate outputs, LLM judges; allows raking of the generation abilities.
  • MC-bench: Request LLMs to build an elaborate structure in Minecraft; outputs can be A/B tested by human judges.

Automated Improvement

See Also

Retrieved from "http://gisaxs.com/index.php?title=AI_Agents&oldid=7564"