language-agents

Tag

Cards List
#language-agents

Communicative On-Policy Distillation: Distilling Black-Box Teacher Capabilities through Communication for Long-Horizon Tasks

arXiv cs.LG ↗ · 2d ago Cached

The paper introduces Communicative On-Policy Distillation (COPD), which uses natural-language communication to distill black-box teacher capabilities into smaller student models for long-horizon tasks, achieving 2.46x the gain of the strongest black-box OPD baseline on CAMG-II with generalization to Terminal-Bench 2.1 and SWE-Bench Pro V2.

0 favorites 0 likes
#language-agents

OfficePrism: Benchmarking Agents on Single-Domain Operations Across Enterprise Business Functions

arXiv cs.CL ↗ · 3d ago Cached

OfficePrism introduces a benchmark of 420 single-domain enterprise tasks across seven business functions, evaluating language agents with executable reference witnesses; across 124 model-harness-package configurations, pass rates range from 22.6% to 76.4%, with performance decreasing from L1 to L3 task complexity.

0 favorites 0 likes
#language-agents

Skill^{infty}: Recursive Self-Improvement for Agent Skill Optimization

Hugging Face Daily Papers ↗ · 2026-10-26 Cached

Skill∞ 介绍了一种递归自我改进框架,让元智能体通过可复用的元技能同时优化任务技能与优化流程本身,在 83 个 SkillsBench 任务上达到 60.7% 的通过率,优于人工整理技能和 SkillEvolver。

0 favorites 0 likes
#language-agents

Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents

Hugging Face Daily Papers ↗ · 2026-10-25 Cached

该论文提出 MICA 与 DART 框架,研究在可执行能力相同的情况下,MCP 与 CLI 接口表示差异如何显著改变语言智能体的行为,并通过针对性修复接口实现提升性能、降低交互成本。

0 favorites 0 likes
#language-agents

Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams

Hugging Face Daily Papers ↗ · 2026-10-19 Cached

论文提出 GRSE(Group-Relative Skill Evolution),通过重复执行并对比同一技能库下的多次运行,将共享团队奖励正确归因到具体的 agent-技能对上,从而改进协作式语言智能体的技能演化,在 HotpotQA 与 LoCoMo 上超越 MASkills 并能从有害技能中恢复。

0 favorites 0 likes
#language-agents

Learning When to Withdraw: Curriculum Coverage and Rule-Specific Failures in Fine-Tuned Language Agents

arXiv cs.AI ↗ · 2026-10-16 Cached

该论文研究了工具使用型语言智能体在监督微调下的「选择性撤回」行为,提出一种能发现聚合准确率所掩盖的规则级回归的诊断方法,并表明微调本身并不足以保证智能体安全。

0 favorites 0 likes
#language-agents

PEARL: Process-Experience Aware Reinforcement Learning for Multi-Turn Agents

arXiv cs.AI ↗ · 2026-10-15 Cached

PEARL提出了一种面向多轮智能体的强化学习方法,通过分支轨迹的首次分歧点构建局部决策特征,并利用成功/失败记忆聚类识别重复模式,从而在不引入额外critic的情况下改进信用分配。在AppWorld和ALFWorld基准上,PEARL在4B和8B规模模型上均超越了ARPO等基线,ALFWorld 4B达到97.9%成功率。

0 favorites 0 likes
#language-agents

Co-Evolving Acting and Forecasting for Language Agents

Hugging Face Daily Papers ↗ · 2026-10-15 Cached

CoPE 论文提出通过在成功轨迹上训练多步动作目标,使语言智能体的行动与预测能力协同演化,从而在不增加推理开销的情况下提升样本效率,相比 GRPO 与 ECHO 在多项基准上取得更强表现。

0 favorites 0 likes
#language-agents

Environmental Feedback Modeling Matters: Rethinking Feedback Treatment in Agentic Hindsight Self-Distillation

arXiv cs.AI ↗ · 2026-10-09 Cached

该论文提出 agentic SElf-distilLation with environmental Feedback modeling (SELF) 框架,将环境反馈建模与后见自蒸馏联合优化,以在缺乏可验证奖励的交互环境中提升语言智能体的训练效果,在 τ-bench 和 AppWorld 基准上显著优于 SDPO 和 GRPO。

0 favorites 0 likes
#language-agents

When History Fails to Become Experience: Action Calibration in Language Agents

arXiv cs.CL ↗ · 2026-10-05 Cached

This paper shows that language agents fail to reliably connect past actions with their outcomes when using interaction history, and proposes a simple outcome-labeling annotation plus a learned calibrator that selectively records experience, improving task success and reducing action repetition.

0 favorites 0 likes
#language-agents

Heavy-Tailed Memory Traces in Long-Horizon Language Agents

arXiv cs.AI ↗ · 2026-10-02 Cached

The paper audits the shape of memory use in long-horizon language agents, finding heavy-tailed (core-tail) retrieval traces, and proposes Core-Tail World Model (CTWM), a rank-based memory controller that cuts prompt tokens by up to 24.48% on LongMemEval while preserving accuracy.

0 favorites 0 likes
#language-agents

NAQD Env: A benchmark for selective withdrawal in language agents

arXiv cs.AI ↗ · 2026-10-01 Cached

NAQD-Env is a synthetic benchmark that evaluates how language agents selectively suspend actions when evidence, permissions, or stop instructions change, finding near-zero withdrawal recall across models despite fine-tuning gains in decision accuracy.

0 favorites 0 likes
#language-agents

RideWay: Benchmarking Efficient Task Completion for Tool-Using Language Agents

arXiv cs.AI ↗ · 2026-09-17 Cached

RideWay introduces an efficiency-centered benchmark and the Efficiency Utility metric for evaluating tool-using language agents in ridehailing tasks, measuring success-gapped performance based on tool calls and user turns.

0 favorites 0 likes
#language-agents

What Should an Agent Forget? Separating What Is Stored from What Is Used

arXiv cs.AI ↗ · 2026-09-11 Cached

This paper introduces RD-Forget, a training-free framework for persistent language agents that separates stored memory from query-conditioned evidence to handle changing facts while preserving historical information.

0 favorites 0 likes
#language-agents

Procedural Memory Under Change: Reuse and Interference in Controlled Web Tasks

arXiv cs.AI ↗ · 2026-09-11 Cached

This study investigates how procedural memory mismatch in language agents affects web tasks, finding that under controlled conditions, mismatch does not lead to behavioral disruption or errors.

0 favorites 0 likes
#language-agents

ContractEval: Query-Conditioned Execution Matching for Procedural Instruction Conformance

arXiv cs.AI ↗ · 2026-09-11 Cached

ContractEval is a diagnostic framework that evaluates procedural conformance in LLM agents by making active obligations explicit and matching them against evidence, detecting failures missed by traditional judges.

0 favorites 0 likes
#language-agents

Outcome Monitors: Recovery Affordances for Silent Tool Failures

arXiv cs.AI ↗ · 2026-08-21 Cached

This paper introduces Outcome Monitors, a deterministic detector that identifies silent failures in tool calls for language agents and provides recovery receipts, significantly improving task completion rates in benchmarks like ToolMaze and τ-bench.

0 favorites 0 likes
#language-agents

AQuA: Recursively Self-Improving Quantitative Trading Research Agents

arXiv cs.CL ↗ · 2026-08-14 Cached

AQuA is a research system with two independent language-model-driven agents that recursively self-improve in quantitative trading research, achieving strong information coefficients on crypto and US equities while using sealed sandboxes to prevent data leakage.

0 favorites 0 likes
#language-agents

Diagnosis Before Recovery: Turning Agent Failures into Selective Self-Correction

arXiv cs.CL ↗ · 2026-08-13 Cached

This paper introduces DARC, a diagnosis-guided recovery harness that makes agent self-correction selective by profiling failure modes and pruning mismatched interventions before test-time correction, improving performance on ALFWorld, AppWorld, and XBRL Finance.

0 favorites 0 likes
#language-agents

EvoGraph-Mem: Failure-Aware Editable Graph Memory for Long-Term Language Agents

arXiv cs.AI ↗ · 2026-08-13 Cached

Presents EvoGraph-Mem, a failure-aware editable graph memory framework for long-term language agents that tracks positive/negative evidence and activation states for insights, enabling memory maintenance through utility-aware retrieval and graph-level editing.

0 favorites 0 likes
Next →
← Back to home

Submit Feedback