ReAuSE integrates a built-in autoregressive search engine into a multi-modal LLM, achieving 9.6% improvement on knowledge-based VQA.

Xinwei Long, Zhiyuan Ma, Ermo Hua et al.

Proposes CodeCriticBench, a comprehensive benchmark for large language models' code critique, combining basic and fine-grained multi-task evaluation with 4,300 samples.

Alexander Zhang, Marcus Dong, Jiaheng Liu et al.

Betti curves derived from persistent homology compress electron density into topological features, improving classification accuracy by 33% over raw data.

Nathan J. Szymanski, Alexander Smith, Prodromos Daoutidis et al.

Establishes the unique support structure of the entropic optimal transport plan in high dimensions with Euclidean cost, supported on transport rays, via relative entropy minimization.

Shrey Aryan, Promit Ghosal

CCGS integrates pointmap association and piecewise-plane constraint to achieve consistent and compact 3D Gaussian segmentation fields, excelling on ScanNet and Replica datasets.

Wenhao Hu, Wenhao Chai, Shengyu Hao et al.

ALFA framework decomposes question quality into attributes, synthesizes data, and uses preference optimization to reduce diagnostic errors by 56.6%.

Shuyue Stella Li, Jimin Mun, Faeze Brahman et al.

Introduces FUR framework using parameter unlearning to evaluate reasoning faithfulness; experiments show effective identification of key reasoning steps.

Martin Tutek, Fateme Hashemi Chaleshtori, Ana Marasović et al.

This paper reviews the evolution of IR architectures, emphasizing transformer-based models and large language models, with performance metrics like nDCG@10 reaching 0.45 on MS MARCO.

Zhichao Xu, Fengran Mo, Zhiqi Huang et al.

This paper introduces PLDM, a latent dynamics planning method that leverages reward-free offline trajectories, outperforming model-free RL in generalization and trajectory stitching.

Vlad Sobal, Wancong Zhang, Kyunghyun Cho et al.

HippoRAG 2 leverages enhanced Passage integration and online LLM optimization, achieving 7% improvement in associative memory and superior factual and sense-making capabilities over standard RAG.

Bernal Jiménez Gutiérrez, Yiheng Shu, Weijian Qi et al.

Logic-RL employs rule-based reinforcement learning trained on synthetic logic puzzles, achieving 125% improvement on AIME and 38% on AMC benchmarks with only 5K puzzles.

Tian Xie, Zitian Gao, Qingnan Ren et al.

EAGER-LLM integrates exogenous behavior and semantic signals via knowledge-rich indices and multi-scale alignment, boosting large language model recommendation performance.

Minjie Hong, Yan Xia, Zehan Wang et al.

Llamba uses cross-architecture distillation from Llama-3.x, achieving 2-3x inference throughput with only 0.1% training data, suitable for edge devices.

Aviv Bick, Tobias Katsch, Nimit Sohoni et al.

Introduced SUnsET dataset to enhance evidence citation accuracy in long-context query summarization.

Dustin Wright, Zain Muhammad Mujahid, Lu Wang et al.

Ensemble perspective on SNN; membrane potential smoothing + guidance boosts accuracy to 83.2% on CIFAR10-DVS with 4 timesteps.

Yongqi Ding, Lin Zuo, Mengmeng Jing et al.

Proposes education-inspired generative and explanation-based MCQA formats, addressing leakage, bias, and difficulty saturation.

Nishant Balepur, Rachel Rudinger, Jordan Lee Boyd-Graber

Using AlphaZero and RNNs for strategic decision-making in Hex and Counter Wargames, showing promising results across terrains.

Guilherme Palma, Pedro A. Santos, João Dias

Proposed an SNN-driven multimodal human action recognition framework, significantly improving energy efficiency and accuracy.

Naichuan Zheng, Hailun Xia, Zeyu Liang et al.

MoVer employs first-order logic for automated spatio-temporal verification of motion graphics, significantly improving correctness in LLM-based synthesis.

Jiaju Ma, Maneesh Agrawala

Enhanced long-context understanding via supervised CoT reasoning; GPT-4o-mini w/ PAI improves 20% on Loong benchmark.

Jingyang Lin, Andy Wong, Tian Xia et al.