Posters

Kangaroo: Double early exiting boosts LLM speed!

KALM: Knowledgeable agents learn complex tasks from LLMs via offline RL using imaginary rollouts, significantly outperforming baselines.

Kaleidoscope: Learnable Masks for Heterogeneous MARL achieves high sample efficiency and policy diversity by using learnable masks for adaptive partial parameter sharing.

Kaleido Diffusion boosts the diversity of images generated by diffusion models without sacrificing quality, using autoregressive latent modeling to add more control and interpretability to the image g…

Boost 3D object detection accuracy by augmenting pseudo-LiDAR point clouds!

I2M2: A novel framework revolutionizes multi-modal learning by jointly modeling inter- and intra-modality dependencies, achieving superior performance across diverse real-world datasets.

Faster John ellipsoid computation achieved via lazy updates and fast matrix multiplication, improving efficiency and enabling low-space streaming algorithms.

JiuZhang3.0 efficiently enhances LLMs’ mathematical reasoning by training a small model to synthesize high-quality training data, drastically reducing costs.

New benchmark and jailbreak method exposes vulnerabilities of LLM moderation, achieving significantly higher success rates than existing methods.

IWBVT: A novel instance weighting approach significantly improves model quality in crowdsourcing by mitigating the impact of intractable instances and achieving a bias-variance trade-off.

iVideoGPT: A scalable, interactive world model trained on millions of human & robot manipulation videos, enabling efficient video prediction and model-based reinforcement learning.

IsoNet++ iteratively refines subgraph matching via early interaction GNNs and node-pair partner interactions, significantly boosting graph retrieval accuracy.

Iteratively Refined Behavior Regularization boosts offline reinforcement learning by iteratively refining the reference policy, ensuring robust and effective control policy learning.

Iterative Reasoning Preference Optimization boosts large language model reasoning by iteratively refining preferences between generated reasoning steps, achieving significant accuracy gains on benchma…

This paper proposes a novel framework, the locally evolving set process, to develop faster localized iterative methods for solving large-scale graph problems, achieving significant speedup over existi…

Researchers reveal how Chain-of-Thought reasoning emerges in transformers via specialized ‘iteration heads’, improving LLM performance and offering insights into mechanistic interpretability.

Offline RL’s performance often lags behind imitation learning, but this paper reveals that policy learning and generalization, not value function learning, are often the main bottlenecks.

New criterion maximizes remaining discrepancy after rationale removal, treating spurious features as noise, improving rationale extraction.

Weighted score matching offers a consistent, efficient solution for estimating parameters in point processes, overcoming the limitations of previous methods.

Large Language Models (LLMs) surprisingly improve the challenging task of Programming by Example (PBE) when fine-tuned on problem-specific data, outperforming classic symbolic methods and even surpass…