Wei Lu

HumanForge reveals that existing benchmarks fall short in capturing the complexities of human interactions in video forensics, exposing critical gaps in current detection capabilities.

DT-Guard outperforms larger models in safety classification while maintaining low-latency performance, proving that reasoning supervision can be efficiently internalized.

RaBitQCache accelerates long-context LLM inference while cutting memory I/O by intelligently adapting token budgets based on attention sparsity.

Ling-2.6 and Ring-2.6 achieve unprecedented efficiency in agentic intelligence, enabling instant responses and deep reasoning at trillion-parameter scale.

Paraphrased prompts may seem equivalent, but they can drastically alter learning dynamics, with some prompts leading to better generalization and less forgetting.

GAIS enables the synthesis of complex tasks that not only outperform existing methods but also do so with dramatically improved data efficiency.

Coordinated attacks on discourse-level opinion can shift perceptions across multi-topic queries, exposing critical vulnerabilities in RAG systems.

Stop letting bad actions ruin your VLA rollouts: Pre-VLA uses runtime verification to boost success rates by 7% while slashing execution steps.

Achieve state-of-the-art shadow removal in remote sensing images without paired training data by unifying shadow detection and removal into a single framework.

Collaboration boosts contextual optimization: sharing beliefs across heterogeneous clients slashes regret in real-world design tasks.

By aligning visual and textual reasoning with human preferences, MARE offers a more accurate and explainable approach to deepfake detection, moving beyond simple classification.