Papers

Field

Top tools here

All tools

Agno5 releases this week
FastMCP6 releases this week
Pydantic AI+17% downloads · 1.5M/wk+17.0%
n8n+25% downloads · 148.3k/wk+25.0%
smolagents+19% downloads · 125.9k/wk+18.7%

Top tools here

Open problems in Agents

Long-horizon task execution and memory3 papers raise it · 3 address it
Agents fail to retain, retrieve, and use past information over long interactions, limiting real-world deployment.
- RealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World Conversationsraises
- Scaling Agentic Data for Long-Horizon Terminal Intelligenceraises · addresses
- Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?raises · addresses
- Agent-Reflex: Teaching Language Agents to Act Reflectivelyaddresses
Real-world task ambiguity and unstated constraints3 papers raise it · 2 address it
Agents struggle with ambiguous, incomplete, or evolving requirements in live systems, unlike controlled benchmarks.
- OR-License: Evaluating LLM Agents as Real-World Operations Research Practitionersraises · addresses
- FDE-Gym: The Last Mile of AI Deliveryraises · addresses
- AgentSWE: Can Coding Agents Build the Agent You Actually Want?raises
Faithful agent development from requirements1 paper raises it · 1 addresses it
Coding agents often produce systems that appear correct but fail to meet the actual requirements.
- AgentSWE: Can Coding Agents Build the Agent You Actually Want?raises · addresses
Sequential distraction in multimodal GUI agents1 paper raises it · 1 addresses it
Omni-modal inputs cause agents to focus on irrelevant signals, leading to compounding errors.
- Focused Omni-Modal GUI Agentsraises · addresses
Open, reproducible agent stacks and evaluation1 paper raises it · 1 addresses it
Closed systems and lack of open benchmarks hinder systematic study and improvement of agent designs.
- OpenGCUA: Toward a Fully Open Stack for Generalist Computer Useraises · addresses

Named by an AI from these papers' abstracts.

Open problems

Recently funded

Market

FirstWorkSeries A · Oct 21$15M
Asymptote LabsSeed · Oct 20$5.5M
LumilabPre-seed · Oct 20undisclosed
GreptileSeries B+ · Oct 19$50M
VesslaPre-seed · Oct 19$3.3M

Recently funded

Topics in this field

coding agents · code generation · coding agent303 papers89 this week
skills · tasks · ai315 papers60 this week
recursive self improvement · harness evolution · adaptive271 papers50 this week
tool using agents · agent systems · multi agent216 papers45 this week
ai agents · indirect prompt injection · adversarial319 papers43 this week
agent · agentic · llm agents172 papers42 this week

Topics in this field

Active contributors · 90 days

Active contributors

Activity · 12 weeks

Activity

Recent papers

All papers in Research

arXiv cs.AIOct 15
Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows
Shirley Anugrah Hayati, Patricia Anugrah Setiani, Ruizi Wang, Zhouyi Li, Khanh Chi Le +2
When humans collaborate with AI agents on long-horizon tasks, what happens to the agents' contributions? For solving long-horizon tasks, humans place greater importance on ongoing interaction with AI agents (e.g., revising plans) than on an agent's ability to complete the task in a single attempt. However, most agentic benchmarks have focused more on automation. Strong performance on such benchmarks does not imply effective real-world collaboration. Professional workflows offer a useful setting, as they often involve interdependent steps. To address this gap, we introduce Kolab, a new dataset of 115+ hours of screen recordings of 50 professionals from 40 occupations collaborating with agents on realistic, long-horizon problems. Kolab also contains participants' collaboration and output quality judgments. Our analysis reveals diverse collaboration patterns, highlighting the importance of when and how professionals coordinate with agents. A collaborative agent must do two things that automation benchmarks never test: anticipate whether experts will adopt, modify, or reject its contribution once it leaves the chat window to guide its next steps, and assess collaboration quality from…
PDF arXiv
arXiv cs.CLOct 15
ToolBottleneckBench: Measuring Tool-Agent Performance under Controlled Pressures and Fixed Harness Budgets
Zichuan Xu, Yuyang Dong, Makoto Onizuka, Chuan Xiao
The success score of a tool agent reflects both its model and the budgets of the harness that runs it. We present ToolBottleneckBench, a benchmark that measures tool-agent performance under controlled pressures and fixed harness budgets. Synthetic enterprise tasks stay fixed while one pressure varies: the registered tool catalog, the visible tool menu, the certified workflow, or the encoding of returned records. An exact grader records the first unmet obligation of each failed episode and the harness limit, if any, that ended it. Across 21 model and harness configurations, the visible menu is the only one of the three pressures that act on the model with a large penalty, and this penalty tracks truncation at the completion cap. Raising the cap sixteenfold removes most of the penalty in the truncation-heavy configurations. Most failures end at a harness limit, and the failure record identifies the assigned pressure no more reliably than a count of tool calls. The repairs that help change which tools reach the model or who issues the calls, and only two repairs, neither improving model behavior, beat every configuration switch. We recommend reporting pressure effects with the…
PDF arXiv
arXiv cs.AIOct 15
Probe-and-Resume: When and How to Evaluate Agent Execution
Jiakai Xu, Tiancheng Ge, Tujie Guan, Yusheng Li, Rohan Timmaraju +2
Test-time search can improve agent performance by exploring alternative actions and using a critic to select promising branches. We study this approach in terminal-based computer-use environments, which expose only a partial view of a large underlying state, and find that the Monte Carlo tree search (MCTS) configuration we evaluate does not significantly improve task success. We identify two sources of poor critic performance: the agent's observations omit environment state needed for evaluation, and intermediate states offer limited evidence of whether a trajectory will succeed. At intermediate states, critics rank continuations only modestly better than chance, and environment access improves this little. We study when to invoke a critic and what access to give it to evaluate a node. Our findings motivate a policy that invokes the critic only at nodes the agent believes are final, rather than during partial progress. Further, the policy gives the critic read, write, and execute access to a fork of the computer's current and initial state, allowing it to probe and perform potentially destructive actions to surface any hidden state the trajectory omits. We call the resulting…
PDF arXiv
arXiv cs.AIOct 15▲ 14
Sci$Π$vot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents
Youran Sun, Xingyu Ren, Chugang Yi, Kejia Zhang, Di Wu +13
Large language model (LLM) agents can keep scientific experiments running for months with little human supervision, but reliable autonomous research also requires scientific judgment. Evaluating scientific judgment is difficult because research decisions are open-ended and depend on project context. We construct Sci$Π$vot, a 248-case benchmark of scientific judgment based on human corrections. The benchmark draws on records from six months of research in which senior researchers ran LLM agents on 26 projects across six scientific fields and repeatedly intervened to correct the agents' research judgments. In each case, an agent chooses between the original LLM plan and the human-corrected alternative, using only the project record available before the correction as context. We compare LLM and human performance and examine the effects of reasoning effort, access to the project record, and reconsideration. At medium reasoning effort with the project record, models systematically prefer the original LLM plan over the human-corrected alternative. Four of five frontier models perform below chance on the selected cases. The highest accuracy in the project condition across the tested…
PDF arXiv
arXiv cs.AIOct 15
OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics
Gideon Shaked, Siddharth Nagisetty
Computational biologists increasingly delegate substantial analyses to AI agents in long-horizon, multi-turn sessions, yet most benchmarks of agents' bioinformatics capabilities pose a single short task with a single-shot prompt. We present OmicsBench, a long-horizon, multi-turn bioinformatics benchmark for AI agents. Each of its 97 tasks links dependent subtasks into a directed acyclic graph, for 662 underlying subtasks in all. Each subtask is posed as one user turn and graded deterministically against an answer key, with optional in-context feedback on incorrect answers. Across 12 models on three harnesses, the slowest agent averages 3.6 hours per task, and no combination comes near saturating the benchmark, with mean scores of 0.43 to 0.57 out of a possible 1 without in-context feedback and 0.54 to 0.67 with it. Multi-turn evaluation also uncovers behavior that single-shot prompts cannot, as agents score lower on subtasks that depend on earlier subtasks, and in-context feedback helps agents recover some of that loss. Because deterministic scoring gives no qualitative account of how an agent failed, we also introduce an unsupervised failure mode discovery pipeline that…
PDF arXiv
arXiv cs.SEOct 15▲ 3
Skip the Compressor: Coding Agents Fix More Issues When They Pick Whole Files
Xiaocong Yang, Rishit Chatterjee, Himanshu Udupi, Vikram Ramavarapu, ChengXiang Zhai
A coding agent works on repositories that can be far larger than its context window can hold. Prompt compressors and retrievers offer solutions to this problem, but patching an issue needs two things at once: the files that must change, and their code exactly as written, since the model must reproduce the lines it edits. Retrievers rank code by its similarity to the issue rather than by whether it must change, and compressors delete text to fit the budget. We show that at an equal token budget, an agent resolves more issues when the budget goes to choosing which files to show verbatim, as our simple selector CQ2-CARE does, than when it goes to pruning tokens or retrieving chunks. We fit every context to the same per-issue budget on 300 recent SWE-rebench issues and grade each patch with the benchmark's tests. With GPT-5.6 and Claude Opus 5, CQ2-CARE resolves 43.7% and 46.7% of the issues, 6.3 points more than the best of three compressors with either model, and 9.3 and 12.7 points more than the best of three retrievers. CQ2-CARE even resolves more issues than two methods trained for issue resolution, SweRank (by 0.7 and 1.7 points) and SWE-Pruner (by 2.0 and 7.3). Our analysis…
PDF arXiv

Papers

Top tools here

All tools

Agno5 releases this week
FastMCP6 releases this week
Pydantic AI+17% downloads · 1.5M/wk+17.0%
n8n+25% downloads · 148.3k/wk+25.0%
smolagents+19% downloads · 125.9k/wk+18.7%

Recently funded

Market

FirstWorkSeries A · Oct 21$15M
Asymptote LabsSeed · Oct 20$5.5M
LumilabPre-seed · Oct 20undisclosed
GreptileSeries B+ · Oct 19$50M
VesslaPre-seed · Oct 19$3.3M

Topics in this field

coding agents · code generation · coding agent303 papers89 this week
skills · tasks · ai315 papers60 this week
recursive self improvement · harness evolution · adaptive271 papers50 this week
tool using agents · agent systems · multi agent216 papers45 this week
ai agents · indirect prompt injection · adversarial319 papers43 this week
agent · agentic · llm agents172 papers42 this week

Active contributors · 90 days

Recent papers

All papers in Research

arXiv cs.AIOct 15
Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows
Shirley Anugrah Hayati, Patricia Anugrah Setiani, Ruizi Wang, Zhouyi Li, Khanh Chi Le +2
When humans collaborate with AI agents on long-horizon tasks, what happens to the agents' contributions? For solving long-horizon tasks, humans place greater importance on ongoing interaction with AI agents (e.g., revising plans) than on an agent's ability to complete the task in a single attempt. However, most agentic benchmarks have focused more on automation. Strong performance on such benchmarks does not imply effective real-world collaboration. Professional workflows offer a useful setting, as they often involve interdependent steps. To address this gap, we introduce Kolab, a new dataset of 115+ hours of screen recordings of 50 professionals from 40 occupations collaborating with agents on realistic, long-horizon problems. Kolab also contains participants' collaboration and output quality judgments. Our analysis reveals diverse collaboration patterns, highlighting the importance of when and how professionals coordinate with agents. A collaborative agent must do two things that automation benchmarks never test: anticipate whether experts will adopt, modify, or reject its contribution once it leaves the chat window to guide its next steps, and assess collaboration quality from…
PDF arXiv
arXiv cs.CLOct 15
ToolBottleneckBench: Measuring Tool-Agent Performance under Controlled Pressures and Fixed Harness Budgets
Zichuan Xu, Yuyang Dong, Makoto Onizuka, Chuan Xiao
The success score of a tool agent reflects both its model and the budgets of the harness that runs it. We present ToolBottleneckBench, a benchmark that measures tool-agent performance under controlled pressures and fixed harness budgets. Synthetic enterprise tasks stay fixed while one pressure varies: the registered tool catalog, the visible tool menu, the certified workflow, or the encoding of returned records. An exact grader records the first unmet obligation of each failed episode and the harness limit, if any, that ended it. Across 21 model and harness configurations, the visible menu is the only one of the three pressures that act on the model with a large penalty, and this penalty tracks truncation at the completion cap. Raising the cap sixteenfold removes most of the penalty in the truncation-heavy configurations. Most failures end at a harness limit, and the failure record identifies the assigned pressure no more reliably than a count of tool calls. The repairs that help change which tools reach the model or who issues the calls, and only two repairs, neither improving model behavior, beat every configuration switch. We recommend reporting pressure effects with the…
PDF arXiv
arXiv cs.AIOct 15
Probe-and-Resume: When and How to Evaluate Agent Execution
Jiakai Xu, Tiancheng Ge, Tujie Guan, Yusheng Li, Rohan Timmaraju +2
Test-time search can improve agent performance by exploring alternative actions and using a critic to select promising branches. We study this approach in terminal-based computer-use environments, which expose only a partial view of a large underlying state, and find that the Monte Carlo tree search (MCTS) configuration we evaluate does not significantly improve task success. We identify two sources of poor critic performance: the agent's observations omit environment state needed for evaluation, and intermediate states offer limited evidence of whether a trajectory will succeed. At intermediate states, critics rank continuations only modestly better than chance, and environment access improves this little. We study when to invoke a critic and what access to give it to evaluate a node. Our findings motivate a policy that invokes the critic only at nodes the agent believes are final, rather than during partial progress. Further, the policy gives the critic read, write, and execute access to a fork of the computer's current and initial state, allowing it to probe and perform potentially destructive actions to surface any hidden state the trajectory omits. We call the resulting…
PDF arXiv
arXiv cs.AIOct 15▲ 14
Sci$Π$vot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents
Youran Sun, Xingyu Ren, Chugang Yi, Kejia Zhang, Di Wu +13
Large language model (LLM) agents can keep scientific experiments running for months with little human supervision, but reliable autonomous research also requires scientific judgment. Evaluating scientific judgment is difficult because research decisions are open-ended and depend on project context. We construct Sci$Π$vot, a 248-case benchmark of scientific judgment based on human corrections. The benchmark draws on records from six months of research in which senior researchers ran LLM agents on 26 projects across six scientific fields and repeatedly intervened to correct the agents' research judgments. In each case, an agent chooses between the original LLM plan and the human-corrected alternative, using only the project record available before the correction as context. We compare LLM and human performance and examine the effects of reasoning effort, access to the project record, and reconsideration. At medium reasoning effort with the project record, models systematically prefer the original LLM plan over the human-corrected alternative. Four of five frontier models perform below chance on the selected cases. The highest accuracy in the project condition across the tested…
PDF arXiv
arXiv cs.AIOct 15
OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics
Gideon Shaked, Siddharth Nagisetty
Computational biologists increasingly delegate substantial analyses to AI agents in long-horizon, multi-turn sessions, yet most benchmarks of agents' bioinformatics capabilities pose a single short task with a single-shot prompt. We present OmicsBench, a long-horizon, multi-turn bioinformatics benchmark for AI agents. Each of its 97 tasks links dependent subtasks into a directed acyclic graph, for 662 underlying subtasks in all. Each subtask is posed as one user turn and graded deterministically against an answer key, with optional in-context feedback on incorrect answers. Across 12 models on three harnesses, the slowest agent averages 3.6 hours per task, and no combination comes near saturating the benchmark, with mean scores of 0.43 to 0.57 out of a possible 1 without in-context feedback and 0.54 to 0.67 with it. Multi-turn evaluation also uncovers behavior that single-shot prompts cannot, as agents score lower on subtasks that depend on earlier subtasks, and in-context feedback helps agents recover some of that loss. Because deterministic scoring gives no qualitative account of how an agent failed, we also introduce an unsupervised failure mode discovery pipeline that…
PDF arXiv
arXiv cs.SEOct 15▲ 3
Skip the Compressor: Coding Agents Fix More Issues When They Pick Whole Files
Xiaocong Yang, Rishit Chatterjee, Himanshu Udupi, Vikram Ramavarapu, ChengXiang Zhai
A coding agent works on repositories that can be far larger than its context window can hold. Prompt compressors and retrievers offer solutions to this problem, but patching an issue needs two things at once: the files that must change, and their code exactly as written, since the model must reproduce the lines it edits. Retrievers rank code by its similarity to the issue rather than by whether it must change, and compressors delete text to fit the budget. We show that at an equal token budget, an agent resolves more issues when the budget goes to choosing which files to show verbatim, as our simple selector CQ2-CARE does, than when it goes to pruning tokens or retrieving chunks. We fit every context to the same per-issue budget on 300 recent SWE-rebench issues and grade each patch with the benchmark's tests. With GPT-5.6 and Claude Opus 5, CQ2-CARE resolves 43.7% and 46.7% of the issues, 6.3 points more than the best of three compressors with either model, and 9.3 and 12.7 points more than the best of three retrievers. CQ2-CARE even resolves more issues than two methods trained for issue resolution, SweRank (by 0.7 and 1.7 points) and SWE-Pruner (by 2.0 and 7.3). Our analysis…
PDF arXiv

Activity · 12 weeks

Fields · AI · updated 1h ago

Agents

LLM agents: planning, tool use, multi-agent systems, computer/web agents and agent benchmarks.

550papers this week0.0% vs last week

All papers · 2.6k With code · 121

Newest Most relevant

2.4k papers this month

Open problems in Agents

Long-horizon task execution and memory3 papers raise it · 3 address it
Agents fail to retain, retrieve, and use past information over long interactions, limiting real-world deployment.
- RealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World Conversationsraises
- Scaling Agentic Data for Long-Horizon Terminal Intelligenceraises · addresses
- Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?raises · addresses
- Agent-Reflex: Teaching Language Agents to Act Reflectivelyaddresses
Real-world task ambiguity and unstated constraints3 papers raise it · 2 address it
Agents struggle with ambiguous, incomplete, or evolving requirements in live systems, unlike controlled benchmarks.
- OR-License: Evaluating LLM Agents as Real-World Operations Research Practitionersraises · addresses
- FDE-Gym: The Last Mile of AI Deliveryraises · addresses
- AgentSWE: Can Coding Agents Build the Agent You Actually Want?raises
Faithful agent development from requirements1 paper raises it · 1 addresses it
Coding agents often produce systems that appear correct but fail to meet the actual requirements.
- AgentSWE: Can Coding Agents Build the Agent You Actually Want?raises · addresses
Sequential distraction in multimodal GUI agents1 paper raises it · 1 addresses it
Omni-modal inputs cause agents to focus on irrelevant signals, leading to compounding errors.
- Focused Omni-Modal GUI Agentsraises · addresses
Open, reproducible agent stacks and evaluation1 paper raises it · 1 addresses it
Closed systems and lack of open benchmarks hinder systematic study and improvement of agent designs.
- OpenGCUA: Toward a Fully Open Stack for Generalist Computer Useraises · addresses

Named by an AI from these papers' abstracts.