A lightweight adapter layer reads hidden generation states from frozen LLMs, reducing requests to larger models by up to 90.7% while maintaining performance.
AI-agent code reviews accelerate code review decisions measurably, but do not improve review quality – a central challenge in automating quality assurance.
SEED leverages self-generated hindsight supervision from language model-native trajectory analysis to bridge the supervision gap between episode-level outcomes and token-level learning signals.
AI agents automate complete attack chains without requiring zero-days, instead systematically exploiting known vulnerabilities and misconfigurations at machine speed.
Qwen-AgentWorld leverages language models as learned environment simulations to efficiently train autonomous agents and improve their reasoning through chain-of-thought prompting.
EDV uses multiple heterogeneous agents to generate diverse solution approaches, an independent verifier, and a consensus mechanism to filter out erroneous experiences before they are…
LLM agents can commit early to an incorrect interpretation without final answer correctness revealing this — hidden-state convergence enables early detection of this failure mode.
RISE achieves similar accuracy to unbounded shell interaction within a limited interaction space, but reduces request costs to about one quarter and scales significantly better…
PaW trains environment models during policy training using the same RL rollouts, consistently improving agent performance without requiring additional simulators or inference costs.
