Natural Language Processing

LLMs struggle to reliably generate valid data science code due to a lack of true understanding of numerical constraints in APIs, despite seemingly mastering common patterns through extensive training.

LLMs struggle with in-context exploration, needing substantial prompt engineering or training interventions to effectively explore multi-armed bandit environments.

LLM agents surprisingly exhibit human-like trust behavior, especially GPT-4, paving the way for simulating complex human interactions in various applications.

LLMs struggle with noisy rationales in chain-of-thought prompting. This paper introduces the NoRa dataset, showing that existing methods struggle. A new method, CD-CoT, significantly improves accura…

Language models learn to skip steps in reasoning, improving efficiency and generalization, showcasing emergent human-like cognitive abilities.

GNNs enhance LLM-based task planning by improving the ability to process task graphs, surpassing existing solutions even without training.

LLMs’ reasoning can be improved by using internal consistency to calibrate their outputs.

Cal-DPO calibrates implicit rewards in contrastive preference learning, dramatically improving large language model alignment with human preferences.

Training large language models efficiently is key; this paper shows how using structured feedforward layers and a novel training regime significantly reduces computational costs and improves training …

AI agents learn human-like communication, combining semantic categorization and pragmatic context-sensitive reasoning, through a novel information-theoretic framework.

Bridge-IF, a novel generative diffusion model, excels at inverse protein folding by learning probabilistic dependencies between protein structures and sequences, significantly outperforming existing m…

SWAB bridges modality and capability gaps in Vision-Language Model selection using optimal transport, enabling accurate prediction of VLM performance without images.

DDSR: a novel sequential recommendation model uses fuzzy sets and discrete diffusion to capture user behavior randomness, outperforming existing methods.

PCNet boosts weakly-supervised referring image segmentation by progressively processing textual cues, mimicking human comprehension, and significantly improving target localization.

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

ViSu boosts semi-supervised scene text recognition by using an online generation strategy for diverse synthetic data and a novel character alignment loss to improve model generalization and robustness…

BoNBON alignment optimizes large language model (LLM) outputs towards human preferences using best-of-n sampling, maximizing win-rate against base models with minimal off-target impact.

BLoB: Bayesian Low-Rank Adaptation by Backpropagation enhances LLMs by jointly tuning mean and covariance of parameters during fine-tuning, improving uncertainty estimation and generalization.

BitDelta drastically shrinks fine-tuned LLMs by quantizing their weight deltas to just one bit, achieving 10x memory reduction and latency improvements without sacrificing performance.

BISCOPE: AI-generated text detection using a novel bidirectional method that outperforms existing techniques by leveraging both prediction and memorization of preceding tokens.