A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Hongjin Qian, Zheng Liu, Peitian Zhang, et al.
Argues long-LLMs are unnecessary for long-context tasks, proposing LC-Boost that enables short-LLMs to solve them via bootstrapping.
Zheyang Xiong, Vasileios Papageorgiou, Kangwook Lee, et al.
Proposes finetuning LLMs on synthetic key-value retrieval data to improve long-context retrieval and reasoning without harming general benchmarks.
Unknown
LongCite enables LLMs to generate fine-grained sentence-level citations in long-context QA, improving trustworthiness via a new benchmark, pipeline, dataset, and trained models.
Unknown
LongLLMLingua enhances LLM performance in long-context tasks via question-aware prompt compression and document reordering.
Unknown
A 137M parameter open-source English text embedding model with 8192 context length outperforming OpenAI on short and long-context tasks.
Unknown
Eagle 2.5 introduces a generalist vision-language model family with Automatic Degrade Sampling and Image Area Preservation for long-context video and high-resolution image understanding.
Unknown
Kimi k1.5 is a multimodal LLM trained with reinforcement learning that achieves state-of-the-art reasoning via long-context scaling and novel long2short compression.
Unknown
Gemini 1.5 Pro is a compute-efficient multimodal mixture-of-experts model excelling in long-context retrieval and understanding across text, video, and audio.
Unknown
GPT-4.5 scales unsupervised learning with new alignment techniques to reduce hallucinations and improve natural conversation, while GPT-4.1 family excels in coding, instruction following, and long-context tasks at lower cost.
Yizhao Gao, Shu-Yu Guo, Shijie Cao, et al.
SeerAttention-R adapts sparse attention for long-context reasoning, achieving strong AIME benchmark results with 4K token budgets and large sparse blocks.
Yizhao Gao, Zhichen Zeng, Dayou Du, et al.
Seerattention introduces a block-sparse attention kernel that achieves a 7.3× speedup at 90% sparsity on 128k sequences, enabling efficient long-context LLMs.
Unknown
LongBioBench uses fictional biographies to controllably test long-context language models, revealing performance degradation beyond 4K tokens.