Leveraging LLMs with prompting and fine-tuning enhances machine translation, especially for low-resource languages and document-level tasks.

Baban Gain, Dibyanayan Bandyopadhyay, Asif Ekbal et al.

Ross3D integrates cross-view and global reconstruction tasks into visual instruction tuning, achieving SOTA with 63.0 EM on SQA3D and 107.0 CIDEr on ScanQA.

Haochen Wang, Yucheng Zhao, Tiancai Wang et al.

PaperBench evaluates AI's ability to autonomously replicate ICML 2024 papers; top model scores 21%.

Giulio Starace, Oliver Jaffe, Dane Sherburn et al.

SpaceR leverages RLVR and map imagination, with 151k data, to enhance video spatial reasoning, outperforming GPT-4o by 11.6%.

Kun Ouyang, Yuanxin Liu, Haoning Wu et al.

This paper reviews the mathematical foundations of proper scoring rules, emphasizing their role in probabilistic estimation and forecast evaluation, with theoretical and practical insights.

Kartik Waghmare, Johanna Ziegel

InfiniteICL breaks context window limits via long short-term memory transformation, reducing context length by 90% and achieving 103% performance.

Bowen Cao, Deng Cai, Wai Lam

Introduces Causal Intervened Likelihood (CIL) for assessing inferability in future forecasting, building the PROPHET benchmark.

Zhengwei Tao, Pu Wu, Zhi Jin et al.

Proposed C2R strategy enhances MoE efficiency by limiting expert collaboration combinations, achieving 0.51% and 0.33% performance gains on LLaMA-MoE and Qwen-MoE.

Mohan Zhang, Pingzhi Li, Jie Peng et al.

Proposes EFDM, an inductive conformal anomaly detection method using elastic distances, effectively identifying shape and magnitude outliers in functional data.

Jason Adams, Brandon Berman, Joshua Michalenko et al.

ffstruc2vec employs a flat, flexible framework for structural node embeddings, enhancing scalability and interpretability in large graphs.

Mario Heidrich, Jeffrey Heidemann, Rüdiger Buchkremer et al.

SuperDec employs superquadric primitives for compact 3D scene decomposition, trained on ShapeNet, generalizes well to ScanNet++ and Replica datasets.

Elisabetta Fedele, Boyang Sun, Leonidas Guibas et al.

Agent S2 employs a hierarchical compositional framework with Mixture-of-Grounding and Proactive Hierarchical Planning, achieving 18.9%-52.8% improvements over SOTA on GUI tasks.

Saaket Agashe, Kyle Wong, Vincent Tu et al.

GenPRM scales test-time compute via generative reasoning, outperforming GPT-4o on ProcessBench.

Jian Zhao, Runze Liu, Kaiyan Zhang et al.

Survey on music generation from single-modal, cross-modal, and multi-modal perspectives, discussing datasets and evaluation methods.

Shuyu Li, Shulei Ji, Zihao Wang et al.

Proposed IMTL-EMLP method enhances learning efficiency via task switching, showing improved accuracy in experiments.

Hanne Say, Suzan Ece Ada, Emre Ugur et al.

DynMoLE boosts MoLE with a hybrid routing mechanism, achieving a 9.6% performance increase.

Dengchun Li, Naizheng Wang, Zihao Zhang et al.

BoundlessBPE breaks pre-tokenization limits, boosting bytes/token by up to 15%.

Craig W. Schmidt, Varshini Reddy, Chris Tanner et al.

Proposes a four-dimensional framework for TTS, including when, how, where, and how well, advancing LLM reasoning capabilities.

Qiyuan Zhang, Fuyuan Lyu, Zexu Sun et al.

KOFFVQA: An objectively evaluated free-form VQA benchmark for Korean language VLMs.

Yoonshik Kim, Jaeyoon Jung

JudgeLRM employs reinforcement learning with outcome-driven rewards to surpass SFT models, with 7B/8B models exceeding GPT-4 in F1 scores.

Nuo Chen, Zhiyuan Hu, Qingyun Zou et al.