Natural Language Processing

SynthDetoxM generates high-quality multilingual parallel data for text detoxification using LLMs, outperforming existing datasets and models in few-shot settings.

Steel-LLM: A fully open-source, resource-efficient Chinese LLM trained with transparency, achieving competitive performance despite limited resources.

ReasonFlux boosts LLM mathematical reasoning by using hierarchical thought templates, outperforming top LLMs like OpenAI’s 01-preview and DeepSeek V3.

Matryoshka Quantization (MatQuant) boosts low-precision model accuracy by up to 10% through a novel multi-scale training approach. It leverages the nested structure of integer data types, allowing a …

Boosting RL fine-tuning efficiency in LLMs: A novel KL penalty modification prioritizes exploration on critical tokens, dramatically improving model performance on arithmetic tasks.

Hephaestus-Forge, a new large-scale pre-training corpus, significantly boosts LLM agent capabilities in API function calling, reasoning, and adaptability through continual pre-training.

OREAL, a novel RL framework, achieves state-of-the-art mathematical reasoning in LLMs using only binary outcome rewards, demonstrating that a 7B model can match the performance of 32B models.

FailSafeQA benchmark rigorously evaluates LLMs’ resilience against diverse human-interaction variations, revealing critical weaknesses in even high-performing models, particularly regarding hallucinat…

Smaller LLMs can outperform larger ones by strategically increasing computation during inference, defying conventional LLM scaling.

Language models learn effective social deduction strategies in a virtual game by using their goal to predict useful information as a dense reward signal, doubling win rates compared to standard RL.

Deep layers in LLMs underperform due to Pre-Layer Normalization; LayerNorm Scaling resolves this by controlling output variance, significantly improving training efficiency.

LM2: Large Memory Models enhance Transformers by adding an auxiliary memory module, significantly improving multi-step reasoning and long-context information synthesis.

APE: a novel method significantly speeds up context-augmented generation (CAG). By using adaptive parallel encoding, APE achieves a 4.5x speedup and maintains high accuracy even with 128K length cont…

Boost LLM reasoning power at test time by recursively processing latent information, enabling dramatic performance gains with fewer parameters.

QuEST enables stable, accurate LLM training using only 1-bit weights and activations, achieving Pareto-optimal performance compared to higher-precision models.

LLMs excel at complex reasoning but struggle with planning; this paper introduces a test-time scaling approach that enhances LLMs’ PDDL reasoning, enabling high-quality PDDL domain generation, outperf…

DuoGuard: a novel two-player RL framework generates high-quality synthetic data, improving multilingual LLM safety by outperforming state-of-the-art models with a significantly smaller model size and …

ARR: A novel zero-shot prompting method significantly boosts LLM performance on diverse question-answering tasks by explicitly incorporating question analysis, information retrieval, and step-by-step …

Simple interactions can easily elicit harmful outputs from LLMs, which are often overlooked. The SPEAK EASY framework and HARMSCORE metric expose this vulnerability and provide tools for better safet…

PILAF optimizes human feedback in reward modeling for better LLM alignment by using a novel response sampling strategy that aligns reward modeling with value optimization.