🏢 Tsinghua University

LLMs can improve reasoning by teaching weaker models, a process called Learning by Teaching (LbT), as shown in this preliminary study. LbT enhances not just student models, but also the teacher model…

C-GAIL stabilizes Generative Adversarial Imitation Learning by applying control theory, resulting in faster convergence, reduced oscillation, and better expert policy matching.

InLine attention, a novel method, bridges the performance gap between softmax and linear attention by incorporating injectivity and local modeling, achieving superior performance while maintaining lin…

New model-agnostic framework for out-of-distribution generalization uses multicalibration across overlapping groups, showing improved robustness and prediction under various distribution shifts.

Boosting LLMs with an intelligent information assistant, ASSISTRAG, significantly improves accuracy and reasoning, especially for less advanced models.

BoostAdapter enhances vision-language model test-time adaptation by combining instance-agnostic historical samples with instance-aware boosting samples for superior out-of-distribution and cross-domai…

Binocular-guided 3D Gaussian splatting with self-supervision generates high-quality novel views from sparse inputs without external priors, significantly outperforming state-of-the-art methods.

AutoTimes repurposes LLMs as autoregressive time series forecasters, achieving state-of-the-art results with minimal trainable parameters and faster training/inference.

iAgents: a novel multi-agent system leveraging LLMs, overcomes information asymmetry by mirroring human social networks to enable effective collaboration in complex tasks, achieving high accuracy in d…

Hypergraph-Based Assembly Fuzzy Representation (HAFR) excels at open-set 3D object retrieval by using part-level shapes and fuzzy representations to overcome challenges posed by unseen object categori…

Mecoin: a novel memory module for efficient graph few-shot class-incremental learning, tackles catastrophic forgetting by employing structured memory units and a memory representation adaptation modul…

AMOR: Adaptable Modular knowledge agent using LLMs, excels with FSM-based reasoning and process feedback, enabling human supervision and domain adaptation.

AlphaTablets revolutionizes 3D planar reconstruction from monocular videos with its novel rectangle-based representation featuring continuous surfaces and precise boundaries, achieving state-of-the-ar…

Efficient Diffusion Alignment (EDA) leverages pretrained diffusion models and Q-functions for efficient continuous control, exceeding all baselines with minimal annotation.

AdaPKC upgrades PeakConv for superior radar semantic segmentation by dynamically adjusting its receptive field, outperforming current state-of-the-art methods.

FlowDPO: Revolutionizing 3D structure prediction with flexible probability paths & Direct Preference Optimization for enhanced accuracy and reduced hallucinations.