Posters

ALPINE reveals how Transformer-based LLMs learn planning by embedding graph information into their weights, but also highlights their inability to handle transitive relationships.

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…

AlphaPruning leverages Heavy-Tailed Self-Regularization theory to allocate optimal layer-wise sparsity ratios in LLMs, achieving 80% sparsity in LLaMA-7B with reasonable perplexity.

AlphaMath: LLMs excel at math reasoning without human-annotated process supervision, using Monte Carlo Tree Search.

Almost-linear RNNs (AL-RNNs) offer highly interpretable symbolic codes for dynamical systems reconstruction, simplifying the analysis of complex systems.

Many graph neural networks (GNNs) surprisingly converge to constant outputs with increasing graph size, limiting their expressiveness.

PSɛBAI+ is a near-optimal algorithm for best arm identification in piecewise stationary linear bandits, efficiently detecting changepoints and aligning contexts for improved accuracy and minimal sampl…

Generalized linear bandits with subexponential reward distributions are self-concordant, enabling second-order regret bounds free of exponential dependence on problem parameters.

New Hallucination-Induced Optimization (HIO) significantly reduces hallucinations in Large Vision-Language Models (LVLMs) by amplifying contrast between correct and incorrect tokens, outperforming exi…

DiMR boosts image generation fidelity by cleverly combining multi-resolution networks with time-dependent layer normalization in diffusion models, achieving state-of-the-art results on ImageNet.

AIMC-CVR: A novel approach that alleviates anchor-shift in incomplete multi-view clustering via cross-view reconstruction, improving accuracy and scalability.

Multi-Path Aggregation (MPA) achieves comparable performance to state-of-the-art methods in multi-task image coding, by unifying feature representations with a novel all-in-one architecture and a two-…

This paper introduces a novel framework for aligning LLMs with honesty, proposing new metrics and training techniques to make LLMs more truthful and less prone to confidently incorrect responses.

This study introduces ’native alignment’ for Arabic LLMs, achieving state-of-the-art results by aligning models during pre-training, rather than post-training.

This paper presents a novel method to align vision models with human aesthetics in image retrieval, using large language models (LLMs) for query rephrasing and preference-based reinforcement learning …

JANUS, a 7B LLM, achieves high alignment to thousands of user preferences by generalizing from diverse system messages, outperforming existing LLMs on various benchmarks.

ALIDIFF aligns target-aware molecule diffusion models with exact energy optimization, generating molecules with state-of-the-art binding energies and improved properties.

Post-processing pre-trained models for alignment using conformal risk control and property testing guarantees better alignment, even when training data is biased.

PRELUDE, a novel framework, leverages user edits of LLM outputs to learn latent preferences, improving agent alignment and minimizing edit costs. CIPHER, its efficient algorithm, infers preferences f…

RE-Control: Aligning LLMs via dynamic representation editing using optimal control theory, achieving superior alignment with significantly fewer resources than fine-tuning.