Harness^2: Recursive Agent Harnessing for an Open World

Harness² improves agent execution harnesses through recursive task-level refinement and continual harness-level learning, boosting performance across benchmarks without ground-truth verifiers.

https://arxiv.org/abs/2610.14224

https://hjfy.top/arxiv/2610.14224

https://github.com/google-research/harness2 (20383)

The Other Half of the Memory Wall: Serving 35B MoEs from SSD with Trained Routing Prediction

Edge0 enables efficient mixture-of-experts inference on consumer hardware by predicting expert routing ahead of time and using recovery adapters to maintain quality with minimal active memory.

https://arxiv.org/abs/2609.18063

https://hjfy.top/arxiv/2609.18063

https://github.com/Edge0-AI/edge0 (7667)

Kronos: A Foundation Model for the Language of Financial Markets

Kronos, a specialized pre-training framework for financial K-line data, outperforms existing models in forecasting and synthetic data generation through a unique tokenizer and autoregressive pre-training on a large dataset.

https://arxiv.org/abs/2508.02739

https://hjfy.top/arxiv/2508.02739

https://github.com/shiyu-coder/Kronos (43956)

MOSS-Image Technical Report

MOSS-Image is an open diffusion transformer that combines high-fidelity synthesis with intent-aware prompt enhancement to generate text-rich, layout-coherent images, achieving state-of-the-art open-weight results on long-form text and layout benchmarks.

https://arxiv.org/abs/2610.23374

https://hjfy.top/arxiv/2610.23374

https://github.com/OpenMOSS/MOSS-Image (2076)

Work Agents: Systems, Capabilities, and Optimization

The paper proposes a unified framework for work agents that integrates foundation models with runtime harnesses to execute complex digital tasks reliably, introducing dimensions of persistence and openness alongside capability families and evaluation metrics.

https://arxiv.org/abs/2610.35845

https://hjfy.top/arxiv/2610.35845

https://github.com/EnjunDu/awesome-working-agent (56)

DynaFlow: Learning Free-Form Multi-Agent Orchestration

DynaFlow enables adaptive multi-agent collaboration by letting a main agent synthesize executable workflows that coordinate sub-agents through structured execution and verification.

https://arxiv.org/abs/2610.34467

https://hjfy.top/arxiv/2610.34467

https://github.com/amiya-special/Dynaflow (53)

Build, Don't Play: Self-Improving LLM Research Agents Develop a Hearthstone Battlegrounds AI That Ri...

AutoTavern demonstrates that self-improving LLM research agents can autonomously build a faithful game simulator and train a reinforcement learning agent that surpasses top human players in Hearthstone Battlegrounds.

https://arxiv.org/abs/2610.33845

https://hjfy.top/arxiv/2610.33845

https://github.com/MCG-NJU/AutoTavern (68)

Zero-Mem: Zero-Token Memory Operations for LLM Agents

Zero-Mem enables structured agent memory without intermediate LLM generation by organizing interaction traces into complementary graph and temporal views for deterministic retrieval.

https://arxiv.org/abs/2607.29377

https://hjfy.top/arxiv/2607.29377

https://github.com/Zero-Mem/Zero-mem (299)

StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative...

StartLux-Decision introduces efficient and adaptive decision models using single-pass prediction and confidence-guided reasoning, trained via a three-stage pipeline with RLCD² reinforcement learning, and evaluates them on Decision Index and the DecisionPivot benchmark.

https://arxiv.org/abs/2610.17309

https://hjfy.top/arxiv/2610.17309

https://github.com/StartLuxLabs/StartLux-Decision (1595)