[Upstream sync] VoltAgent/awesome-ai-agent-papers (github) — 0 added, 1 modified #52
@@ -2,17 +2,15 @@
|
||||
title: "Awesome AI Agent Papers"
|
||||
task: ""
|
||||
lineage_type: import
|
||||
upstream_source: https://github.com/VoltAgent/awesome-ai-agent-papers/blob/b4704753/README.md
|
||||
upstream_sha: b4704753
|
||||
imported_at: 2026-09-04
|
||||
upstream_source: https://github.com/VoltAgent/awesome-ai-agent-papers/blob/6bc94c58/README.md
|
||||
upstream_sha: 6bc94c58
|
||||
imported_at: 2026-10-26
|
||||
prompt_class: catalogue
|
||||
upstream_changes: accepted
|
||||
author: upstream
|
||||
validated: false
|
||||
---
|
||||
|
||||
|
||||
|
||||
<div align="center">
|
||||
|
||||
<a href="https://github.com/VoltAgent/voltagent">
|
||||
@@ -31,18 +29,12 @@ validated: false
|
||||
</div>
|
||||
|
||||
[](https://awesome.re)
|
||||
|
||||
|
||||
|
||||
[](https://s.voltagent.dev/discord)
|
||||
|
||||
</div>
|
||||
|
||||
<div align="center">
|
||||
<strong>More awesome collections for developers</strong>
|
||||
<br />
|
||||
<br />
|
||||
</div>
|
||||
|
||||
|
||||
|
||||
# Awesome AI Agent Papers
|
||||
@@ -72,13 +64,18 @@ Hundreds of papers are published on arXiv every week, and a growing number of th
|
||||
- [Memory & RAG](#memory--rag) (58)
|
||||
- [Multi-Agent](#multi-agent) (54)
|
||||
- [Memory & RAG](#memory--rag) (57)
|
||||
- [Eval & Observability](#eval--observability) (80)
|
||||
- [Eval & Observability](#eval--observability) (87)
|
||||
- [Agent Tooling](#agent-tooling) (95)
|
||||
- [AI Agent Security](#ai-agent-security) (83)
|
||||
- [Agent Tooling](#agent-tooling) (100)
|
||||
- [AI Agent Security](#ai-agent-security) (82)
|
||||
|
||||
<br>
|
||||
|
||||
|
||||
## AI Design + Build Ecosystem Tools
|
||||
|
||||
|
||||
<br/>
|
||||
|
||||
You ship products with AI, but every launch still dies quietly because nobody posts about it. [EveryFeed](https://everyfeed.ai/) plugs your AI assistant into a social workspace that drafts, schedules, and publishes across 35+ channels — no agency, no marketing hire.
|
||||
@@ -87,25 +84,31 @@ You ship products with AI, but every launch still dies quietly because nobody po
|
||||
<img src="https://cdn.voltagent.dev/awesome-repo/everyfeed-social.png" alt="everyfeed" /><br/>
|
||||
</a>
|
||||
|
||||
<br/>
|
||||
<br/>
|
||||
|
||||
Stop building from a blank page. [LaunchKit](https://launchkit.getdesign.md/) gives your AI coding assistant a complete, working product to start from — websites, startups, and web apps that are clickable on day one.
|
||||
|
||||
|
||||
<a href="https://launchkit.getdesign.md/">
|
||||
<img src="https://cdn.voltagent.dev/awesome-repo/new-launchkit.png" alt="launchkit" /><br/>
|
||||
<img src="https://cdn.voltagent.dev/awesome-repo/website-starter-kit-banner-dark-0315e5f9c1.png" alt="launchkit" /><br/>
|
||||
</a>
|
||||
|
||||
<br/>
|
||||
|
||||
|
||||
<a href="https://mobile-starterkit.getdesign.md/">
|
||||
<img src="https://cdn.voltagent.dev/awesome-repo/mobile-starter-kit-banner-light-450ba0a9b0.png" alt="mobilekit" /><br/>
|
||||
</a>
|
||||
|
||||
<br/>
|
||||
|
||||
<details open id="multi-agent">
|
||||
<summary><h3 style="display:inline">Multi-Agent (54)</h3></summary>
|
||||
<summary><h3 style="display:inline">Multi-Agent (55)</h3></summary>
|
||||
|
||||
<br>
|
||||
|
||||
| Paper | arXiv ID |
|
||||
|---|:---:|
|
||||
| **[AgonAlpha: Autonomous Alpha Discovery via Prompt Economy and Scalable Agentic Search](https://arxiv.org/pdf/2608.11250)** - A proposer agent writes each trading factor as a full research record (hypothesis, executable expression, platform results, rationale), and a separate reviewer with fresh context re-runs the evaluation and can veto made-up claims. On WorldQuant BRAIN it got 17 SPECTACULAR-grade alphas out of 60 submissions across five users and six model backends, with Sharpe up to 3.48. | <a href="https://arxiv.org/abs/2608.11250"><img src="https://img.shields.io/badge/arXiv-2608.11250-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Agon: An Autonomous Large-Scale Omnidisciplinary Research System Built on Prompt Economy](https://arxiv.org/pdf/2606.24177)** - Treats prompt engineering as an engineering discipline rather than an art, minimizing human time while expecting maximum output. Carries the smallest prompt footprint among cross-disciplinary autoresearch systems (18 roles, 230.6 KiB total), has run across 10+ research fields without modification, and the longest observed run went 30 days unattended. | <a href="https://arxiv.org/abs/2606.24177"><img src="https://img.shields.io/badge/arXiv-2606.24177-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[PerceptUI: LLM Agents as Human-Aligned Synthetic Users for UI/UX Evaluation](https://arxiv.org/pdf/2606.05697)** - A persona-conditioned framework that predicts how a specific user would answer UI/UX evaluation questions and explains why in natural language. Trained via contrastive reflection fine-tuning and reflective prompt evolution, reaching human-level realism and generalizing to unseen questions and personas. | <a href="https://arxiv.org/abs/2606.05697"><img src="https://img.shields.io/badge/arXiv-2606.05697-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[OpenCLAW-P2P v6.0: Resilient Multi-Layer Persistence, Live Reference Verification, and Production-Scale Evaluation of Decentralized AI Peer Review](https://arxiv.org/pdf/2604.19792)** - Presents a decentralized AI peer-review platform where autonomous agents publish, score, verify references, and preserve research papers across a multi-layer storage and retrieval architecture. | <a href="https://arxiv.org/abs/2604.19792"><img src="https://img.shields.io/badge/arXiv-2604.19792-b31b1b.svg" alt="arXiv" /></a> |
|
||||
@@ -241,12 +244,16 @@ Stop building from a blank page. [LaunchKit](https://launchkit.getdesign.md/) gi
|
||||
<br>
|
||||
|
||||
<details id="eval--observability">
|
||||
<summary><h3 style="display:inline">Eval & Observability (81)</h3></summary>
|
||||
<summary><h3 style="display:inline">Eval & Observability (87)</h3></summary>
|
||||
|
||||
<br>
|
||||
|
||||
| Paper | arXiv ID |
|
||||
|---|:---:|
|
||||
| **[Super Research: Answering Highly Complex Questions with Large Language Models through Super Deep and Super Wide Research](https://arxiv.org/pdf/2603.00582v2)** - Tests research agents on 300 expert-written questions that require planning, broad evidence gathering, and in-depth investigation across sources. The public benchmark includes 3,600 reports from twelve systems for inspecting how they approach these tasks. | <a href="https://arxiv.org/abs/2603.00582v2"><img src="https://img.shields.io/badge/arXiv-2603.00582-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Empowering Edge Agents: A Systematic Analysis of Edge-Cloud Collaboration in LLM Agent Execution](https://arxiv.org/pdf/2610.17816)** - A benchmark of 128 OpenClaw workspace tasks where a small edge model and a cloud model share the work, scored at once on task completion, cloud cost and edge FLOPs, and how much annotated sensitive data reaches the cloud. With Qwen3.5-2B on the edge, per-step routing raises completion from 12.75 to 39.97 while using 22.7% of the cloud-only cost. | <a href="https://arxiv.org/abs/2610.17816"><img src="https://img.shields.io/badge/arXiv-2610.17816-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces](https://arxiv.org/pdf/2609.33295)** - Turns descriptions of unwanted agent behavior into tests built from real deployment logs. Each test scores the model's next response at a recorded decision point, including tool use, error handling, and claims about finished work, without running its actions. | <a href="https://arxiv.org/abs/2609.33295"><img src="https://img.shields.io/badge/arXiv-2609.33295-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Why Do AI Agents Break Rules? How Framing, Context, and Social Signals Shape Compliance](https://arxiv.org/pdf/2608.12323)** - Puts twelve LLMs in a procurement-agent role under one environmental rule and varies how the rule is framed, what the penalty is, and who pushes back. Compliance spans 46 percentage points across models under identical conditions, and stating the fine, manager demands, peer outcomes, and user pressure each raise violation rates. | <a href="https://arxiv.org/abs/2608.12323"><img src="https://img.shields.io/badge/arXiv-2608.12323-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[PerspectiveGap: A Benchmark for Multi-Agent Orchestration Prompting](https://arxiv.org/pdf/2606.08878)** - A 110-scenario benchmark testing whether LLMs can compose orchestration prompts that distribute context to sub-agents without information leakage. Across 10 topologies and 27 commercial models, GPT-5.5 leads with 62% pass rate while the average is 14.9%. | <a href="https://arxiv.org/abs/2606.08878"><img src="https://img.shields.io/badge/arXiv-2606.08878-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[RewardHarness: Self-Evolving Agentic Post-Training](https://arxiv.org/pdf/2605.08703)** - Evolves a library of scoring skills and tools from preference examples, then uses a frozen vision-language sub-agent to evaluate image edits and produce a reward for GRPO training. | <a href="https://arxiv.org/abs/2605.08703"><img src="https://img.shields.io/badge/arXiv-2605.08703-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[ClawBench: Evaluating Browser Agents on Live Production Websites with Submission-Interception](https://arxiv.org/abs/2604.08523)** - Benchmarks browser agents on 283 everyday tasks (V1 153 + V2 130) across 163 live production sites, with a Chrome-extension plus CDP layer that blocks only the final write request so agents can run end-to-end on real sites without real-world side effects. Two-stage scoring (interception + LLM judge); leaderboard at https://claw-bench.com. | <a href="https://arxiv.org/abs/2604.08523"><img src="https://img.shields.io/badge/arXiv-2604.08523-b31b1b.svg" alt="arXiv" /></a> |
|
||||
@@ -337,12 +344,17 @@ Stop building from a blank page. [LaunchKit](https://launchkit.getdesign.md/) gi
|
||||
<br>
|
||||
|
||||
<details id="agent-tooling">
|
||||
<summary><h3 style="display:inline">Agent Tooling (97)</h3></summary>
|
||||
<summary><h3 style="display:inline">Agent Tooling (98)</h3></summary>
|
||||
<summary><h3 style="display:inline">Agent Tooling (100)</h3></summary>
|
||||
|
||||
<br>
|
||||
|
||||
| Paper | arXiv ID |
|
||||
|---|:---:|
|
||||
| **[Jev in the Wild: A Data-Driven Analysis of the Jev Model's Functionality, Applications and Ecosystem](https://arxiv.org/pdf/2609.30216)** - Surveys 2,170 public GitHub Jev projects to show which application domains use typed decisions and how those uses spread during the model's early growth. | <a href="https://arxiv.org/abs/2609.30216"><img src="https://img.shields.io/badge/arXiv-2609.30216-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Steer, Don't Solve: Training Small Critic Models for Large Code Agents](https://arxiv.org/pdf/2606.21811)** - Trains 4B and 8B critic models with SFT and DPO to identify errors in a coding agent's trajectory and give high-level guidance every few steps at inference time, without producing actions. The critics improve SWE-bench Verified resolve rates of six larger agents, by up to 16 points, and lower inference cost for some agents by finishing tasks in fewer steps. | <a href="https://arxiv.org/abs/2606.21811"><img src="https://img.shields.io/badge/arXiv-2606.21811-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Training Agents to Evolve with Their Harness: TaoLive Digital Avatar Agent Technical Report](https://arxiv.org/pdf/2608.15763v3)** - Trains a compact agent while varying its Skills, tool schemas, prompts, and Hooks, so it can handle later harness changes without retraining for each edit. Tests developer-reviewed updates with frozen model weights and reports a live-commerce deployment. | <a href="https://arxiv.org/abs/2608.15763"><img src="https://img.shields.io/badge/arXiv-2608.15763-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Graph of States: Solving Abductive Tasks with Large Language Models](https://arxiv.org/pdf/2603.21250)** - Organizes multi-agent reasoning around a shared causal graph of hypotheses, evidence, confidence scores, and causal links, while a state machine controls when agents investigate, refine, backtrack, or stop. Evaluated on medical and distributed-system diagnosis tasks. |<a href="https://arxiv.org/abs/2603.21250"><img src="https://img.shields.io/badge/arXiv-2607.08983-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[SCATE: Learning to Supervise Coding Agents for Cost-Effective Test Generation](https://arxiv.org/pdf/2607.08983)** - Replaces the human who keeps prodding a coding agent to write more tests with a contextual bandit that picks the next testing action from current coverage and class testability signals. Driving GEMINI-CLI, it reaches 32.3% higher line coverage and 30.9% higher branch coverage than the agent on its own, and it learns a different policy when the same setup drives CLAUDE CODE. | <a href="https://arxiv.org/abs/2607.08983"><img src="https://img.shields.io/badge/arXiv-2607.08983-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Ouroboros: A Self-Developing Frontier Coding Agent with Reviewed Core Evolution](https://arxiv.org/pdf/2608.08311)** - Documents a coding-agent harness that can update its tools, prompts, context assembly, and core code through reviewed commits, with a 161-day live deployment. Reports frozen-snapshot results of 86.74% on Terminal-Bench 2.1 and 90.69% on OSWorld-Verified. | <a href="https://arxiv.org/abs/2608.08311"><img src="https://img.shields.io/badge/arXiv-2608.08311-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[On Effectiveness and Efficiency of Agentic Tool-calling and RL Training](https://arxiv.org/pdf/2606.00135)** - Find that current agentic tool-calling benchmarks like BFCL are quite brittle: system prompt, multi-turn template or even pure random seeds could have huge influence on the final performance. | <a href="https://arxiv.org/abs/2606.00135"><img src="https://img.shields.io/badge/arXiv-2602.06875-b31b1b.svg" alt="arXiv" /></a> |
|
||||
@@ -441,18 +453,20 @@ Stop building from a blank page. [LaunchKit](https://launchkit.getdesign.md/) gi
|
||||
| **[CaveAgent: Transforming LLMs into Stateful Runtime Operators](https://arxiv.org/pdf/2601.01569v3)** - Proposes a dual-stream architecture that elevates the persistent Python runtime as the central locus of agent state, with stateful runtime management and skill injection for long-horizon tasks. | <a href="https://arxiv.org/abs/2601.01569v3"><img src="https://img.shields.io/badge/arXiv-2601.01569-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Actively Obtaining Environmental Feedback for Autonomous Action Evaluation Without Predefined Measurements](https://arxiv.org/pdf/2601.04235v1)** - Proposes an active feedback model where AI agents proactively interact with the environment to discover and verify feedback without relying on predefined measurements. | <a href="https://arxiv.org/abs/2601.04235v1"><img src="https://img.shields.io/badge/arXiv-2601.04235-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Warp-Cortex: An Asynchronous, Memory-Efficient Architecture for Million-Agent Cognitive Scaling on Consumer Hardware](https://arxiv.org/pdf/2601.01298v1)** - Proposes an asynchronous architecture for million-agent scaling that reduces memory complexity via singleton weight sharing and topological synapse-inspired KV-cache sparsification. | <a href="https://arxiv.org/abs/2601.01298v1"><img src="https://img.shields.io/badge/arXiv-2601.01298-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training](https://arxiv.org/pdf/2609.04094)** - Trains long-horizon tool-using agents with RL when no ground-truth success signal is available. Generates task-specific rubrics during training, scores each finished trajectory once with an LLM judge, and redistributes that score over the steps behind each criterion to get per-step GRPO advantages; on AppWorld it gains 15.9 points over the base model and 5.3 over GRPO trained on the ground-truth unit-test reward. | <a href="https://arxiv.org/abs/2609.04094"><img src="https://img.shields.io/badge/arXiv-2609.04094-b31b1b.svg" alt="arXiv" /></a> |
|
||||
|
||||
</details>
|
||||
|
||||
<br>
|
||||
|
||||
<details id="ai-agent-security">
|
||||
<summary><h3 style="display:inline">AI Agent Security (82)</h3></summary>
|
||||
<summary><h3 style="display:inline">AI Agent Security (83)</h3></summary>
|
||||
|
||||
<br>
|
||||
|
||||
| Paper | arXiv ID |
|
||||
|---|:---:|
|
||||
| **[One Polluted Page Is Enough: Evaluating Web Content Pollution in LLM Recommenders](https://arxiv.org/pdf/2606.13610)** - Tests whether search-augmented LLMs recommend brands that do not exist, by swapping the real brand name inside the top-ranked pages the model retrieves and changing nothing else. Across 12 models, one altered page at rank 1 produces a fake recommendation in up to 27% of cases, and altering the top three pages reaches 73.8%. | <a href="https://arxiv.org/abs/2606.13610"><img src="https://img.shields.io/badge/arXiv-2606.13610-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Internal Safety Collapse in Frontier Large Language Models](https://arxiv.org/pdf/2603.23509)** - Reveals that AI agents produce harmful content (toxic text, exploits, dangerous data) as a side effect of completing normal professional tasks — no adversarial prompting needed. At least one mode (single-turn, ICL, or agentic) succeeds on every frontier model tested. 56 cross-domain scenarios across 8+ disciplines. | <a href="https://arxiv.org/abs/2603.23509"><img src="https://img.shields.io/badge/arXiv-2603.23509-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Confundo: Learning to Generate Robust Poison for Practical RAG Systems](https://arxiv.org/pdf/2602.06616v1)** - Trains an LLM to generate RAG poison that survives real-world content processing and query variation for stress-testing RAG defenses. | <a href="https://arxiv.org/abs/2602.06616v1"><img src="https://img.shields.io/badge/arXiv-2602.06616-b31b1b.svg" alt="arXiv" /></a> |
|
||||
| **[Malicious Agent Skills in the Wild: A Large-Scale Security Empirical Study](https://arxiv.org/pdf/2602.06547v1)** - Analyzes 98K agent skills from community registries to study the prevalence and nature of malicious third-party agent plugins. | <a href="https://arxiv.org/abs/2602.06547v1"><img src="https://img.shields.io/badge/arXiv-2602.06547-b31b1b.svg" alt="arXiv" /></a> |
|
||||
|
||||
Reference in New Issue
Block a user