14 papers
A Theory of Reliable Self-Evolution for Agent Harnesses
Qianshu Cai, Yonggang Zhang, Jun Nie +6
In harness self-evolution, agents modify their own prompts, code, tools, and orchestration while keeping the underlying language model fixed. Recent work has shown that agents can…
Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence
Zhiqin Yang, Jingwen Fu, Yuhan Liu +16
Recent advances in large reasoning models (LRMs) have shown that reinforcement learning with verifiable rewards (RLVR) can substantially improve reasoning in mathematics and code,…
EvolveNet: Collaborative Harness Evolution for Agent Self-Improvement
Jun Nie, Yonggang Zhang, Qianshu Cai +3
An LLM agent can acquire persistent improvements by evolving its harness---the executable program governing model calls, tool use, verification, and recovery---without updating mod…
Zero2Skill: Bootstrapping Robot Skills through Autonomous Data Collection, Training, and Deployment
Boyuan Wang, Zhenyuan Zhang, Zhiqin Yang +16
Autonomous data collection governs the volume and quality of real-world trajectories for manipulation policy learning. Existing pipelines reduce human effort via self-resetting, VL…
DRNOISE: Benchmarking Deep Research Agents in Misleading Evidence Environments
Jun Nie, Zhiqin Yang, Zhenheng Tang +4
Deep research agents increasingly operate over the open web, where relevant records coexist with redundant summaries, outdated reports, and misleading documents. Existing evaluatio…
A Control Theory of Predictability in Latent World Models
Hanzhe You, Yonggang Zhang, Maohao Ran +6
Latent world models are trained to predict future states in a learned representation and are then deployed inside a planner that selects actions by simulating them forward. Current…
TTHE: Test-Time Harness Evolution
Jun Nie, Yonggang Zhang, Jun Song +5
The behavior of an LLM agent is determined not only by the underlying model, but also by its harness: the executable program that constructs context, invokes tools, verifies interm…
Scaling Multi-Hop Training Data via Graph-Constrained Path Selection
Pengyu Chen, Yonggang Zhang, Mingming Chen +3
Endowing large language models with compositional reasoning over specialized documents requires multi-hop training data at scale, where such data rarely exists outside of curated b…
MOSS: Self-Evolution through Source-Level Rewriting in Autonomous Agent Systems
Qianshu Cai, Yonggang Zhang, Xianzhang Jia +5
Autonomous agentic systems are largely static after deployment: they do not learn from user interactions, and recurring failures persist until the next human-driven update ships a…
Conditional Equivalence of DPO and RLHF: Implicit Assumption, Failure Modes, and Provable Alignment
Zhiqin Yang, Yonggang Zhang, Wei Xue +3
Direct Preference Optimization (DPO) has emerged as a popular alternative to Reinforcement Learning from Human Feedback (RLHF), offering theoretical equivalence with simpler implem…
ClawNet: Human-Symbiotic Agent Network for Cross-User Autonomous Cooperation
Zhiqin Yang, Zhenyuan Zhang, Xianzhang Jia +4
Current AI agent frameworks have made remarkable progress in automating individual tasks, yet all existing systems serve a single user. Human productivity rests on the social and o…
IFFair: Influence Function-driven Sample Reweighting for Fair Classification
Jingran Yang, Min Zhang, Lingfeng Zhang +2
Because machine learning has significantly improved efficiency and convenience in the society, it's increasingly used to assist or replace human decision-making. However, the data-…
FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning
Zhiqin Yang, Yonggang Zhang, Chenxin Li +3
Federated Learning (FL) confronts a significant challenge known as data heterogeneity, which impairs model performance and convergence. Existing methods have made notable progress…
Semantic Voting: A Self-Evaluation-Free Approach for Efficient LLM Self-Improvement on Unverifiable Open-ended Tasks
Chunyang Jiang, Yonggang Zhang, Yiyang Cai +5
The rising cost of acquiring supervised data has driven significant interest in self-improvement for large language models (LLMs). Straightforward unsupervised signals like majorit…