We present RA VEN ( R ecognition and A daptation of Video ENtities), an adaptive AI agent framework designed for mul-timodal entity discovery and retrieval in large-scale video collections. Synthesizing information across visual, audio, and textual modalities, RA VEN autonomously processes video data to produce structured, actionable representations for downstream tasks. Key contributions include (1) a category understanding step to infer video themes and general-purpose entities, (2) a schema generation mechanism that dynamically defines domain-specific entities and attributes, and (3) a rich entity extraction process that leverages semantic retrieval and schema-guided prompting. RA VEN is designed to be model-agnostic, allowing the integration of different vision-language models (VLMs) and large language models (LLMs) based on application-specific requirements. This flexibility supports diverse applications in personalized search, content discovery, and scalable information retrieval, enabling practical applications across vast datasets.

Large language models (LLMs) excel in question-answering (QA) tasks, and retrieval-augmented generation (RAG) enhances their precision by incorporating external evidence from diverse sources like web pages, databases, and knowledge graphs. However, current RAG methods rely on agent-specific strategies for individual data sources, posing challenges low-resource or black-box environments and complicates operations when evidence is fragmented across sources. To address these limitations, we propose ER-RAG, a framework that unifies evidence integration across heterogeneous data sources using the Entity-Relationship (ER) model. ER-RAG standardizes entity retrieval and relationship querying through ER-based APIs with GET and JOIN operations. It employs a two-stage generation process: first, a preference optimization module selects optimal sources; second, another module constructs API chains based on source schemas. This unified approach allows efficient fine-tuning and seamless integration across diverse data sources. ER-RAG demonstrated its effectiveness by winning all three tracks of the 2024 KDDCup CRAG Challenge, achieving performance on par with commercial RAG pipelines using an 8B LLM backbone. It outperformed hybrid competitors by 3.1% in LLM score and accelerated retrieval by 5.5X.

Misinformation continues to pose a significant challenge in today's information ecosystem, profoundly shaping public perception and behavior. Among its various manifestations, Out-of-Context (OOC) misinformation is particularly obscure, as it distorts meaning by pairing authentic images with misleading textual narratives. Existing methods for detecting OOC misinformation predominantly rely on coarse-grained similarity metrics between image-text pairs, which often fail to capture subtle inconsistencies or provide meaningful explainability. While multi-modal large language models (MLLMs) demonstrate remarkable capabilities in visual reasoning and explanation generation, they have not yet demonstrated the capacity to address complex, fine-grained, and cross-modal distinctions necessary for robust OOC detection. To overcome these limitations, we introduce EXCLAIM, a retrieval-based framework designed to leverage external knowledge through multi-granularity index of multi-modal events and entities. Our approach integrates multi-granularity contextual analysis with a multi-agent reasoning architecture to systematically evaluate the consistency and integrity of multi-modal news content. Comprehensive experiments validate the effectiveness and resilience of EXCLAIM, demonstrating its ability to detect OOC misinformation with 4.3% higher accuracy compared to state-of-the-art approaches, while offering explainable and actionable insights.

As text and code resources have expanded, large-scale pre-trained models have shown promising capabilities in code generation tasks, typically employing supervised fine-tuning with problem statement-program pairs. However, increasing model size and data volume for performance gains also raises computational demands and risks of overfitting. Addressing these challenges, we present RETROcode, a novel adaptation of the RETRO architecture \cite{RETRO} for sequence-to-sequence models, utilizing a large code database as an auxiliary scaling method. This approach, diverging from simply enlarging model and dataset sizes, allows RETROcode to leverage a vast code database for prediction, enhancing the model's efficiency by integrating extensive memory. Our findings indicate that RETROcode not only outperforms similar-sized traditional architectures on test sets but also approaches the effectiveness of the much larger Codex model, despite being trained from scratch on a substantially smaller dataset.

Despite the recent success of large language models (LLMs), LLMs are particularly challenging in long-sequence inference scenarios due to the quadratic computational complexity of the attention mechanism. Inspired by the interpretability theory of feature attribution in neural network models, we observe that not all tokens have the same contribution. Based on this observation, we propose a novel token pruning framework, namely Saliency-driven Dynamic Token Pruning (SDTP), to gradually and dynamically prune redundant tokens based on the input context. Specifically, a lightweight saliency-driven prediction module is designed to estimate the importance score of each token with its hidden state, which is added to different layers of the LLM to hierarchically prune redundant tokens. Furthermore, a ranking-based optimization strategy is proposed to minimize the ranking divergence of the saliency score and the predicted importance score. Extensive experiments have shown that our framework is generalizable to various models and datasets. By hierarchically pruning 65\% of the input tokens, our method greatly reduces 33\% $\sim$ 47\% FLOPs and achieves speedup up to 1.75$\times$ during inference, while maintaining comparable performance. We further demonstrate that SDTP can be combined with KV cache compression method for further compression.

We introduce ShieldGemma 2, a 4B parameter image content moderation model built on Gemma 3. This model provides robust safety risk predictions across the following key harm categories: Sexually Explicit, Violence \& Gore, and Dangerous Content for synthetic images (e.g. output of any image generation model) and natural images (e.g. any image input to a Vision-Language Model). We evaluated on both internal and external benchmarks to demonstrate state-of-the-art performance compared to LlavaGuard \citep{helff2024llavaguard}, GPT-4o mini \citep{hurst2024gpt}, and the base Gemma 3 model \citep{gemma_2025} based on our policies. Additionally, we present a novel adversarial data generation pipeline which enables a controlled, diverse, and robust image generation. ShieldGemma 2 provides an open image moderation tool to advance multimodal safety and responsible AI development.

Parameter generation has emerged as a novel paradigm for neural network development, offering an alternative to traditional neural network training by synthesizing high-quality model weights directly. In the context of Low-Rank Adaptation (LoRA) for evolving ($\textit{i.e.}$, constantly updated) large language models (LLMs), this approach promises efficient adaptation without costly retraining. However, existing methods face critical limitations in simultaneously achieving scalability and controllability. In this paper, we introduce $\texttt{ORAL}$, a novel $\textbf{conditional recurrent diffusion}$ framework that addresses these challenges. $\texttt{ORAL}$ incorporates a novel conditioning mechanism that integrates model architecture and textual task specifications, enabling the generation of task-specific LoRA parameters that can seamlessly transfer across evolving foundation models. Our approach successfully scales to billions-of-parameter LLMs and maintains controllability. Through extensive experiments across seven language tasks, four vision tasks, and three multimodal tasks using five pre-trained LLMs, we demonstrate that $\texttt{ORAL}$ generates high-quality LoRA parameters that achieve comparable or superior performance to vanilla trained counterparts.

Multimodal large language models (MLLMs), such as GPT-4o, are garnering significant attention. During the exploration of MLLM training, we identified Modality Composition Incoherence, a phenomenon that the proportion of a certain modality varies dramatically across different examples. It exacerbates the challenges of addressing mini-batch imbalances, which lead to uneven GPU utilization between Data Parallel (DP) instances and severely degrades the efficiency and scalability of MLLM training, ultimately affecting training speed and hindering further research on MLLMs. To address these challenges, we introduce OrchMLLM, a comprehensive framework designed to mitigate the inefficiencies in MLLM training caused by Modality Composition Incoherence. First, we propose Batch Post-Balancing Dispatcher, a technique that efficiently eliminates mini-batch imbalances in sequential data. Additionally, we integrate MLLM Global Orchestrator into the training framework to orchestrate multimodal data and tackle the issues arising from Modality Composition Incoherence. We evaluate OrchMLLM across various MLLM sizes, demonstrating its efficiency and scalability. Experimental results reveal that OrchMLLM achieves a Model FLOPs Utilization (MFU) of $41.6\%$ when training an 84B MLLM with three modalities on $2560$ H100 GPUs, outperforming Megatron-LM by up to $3.1\times$ in throughput.

A group of organizations, including nonprofits like LatinoProsperity and labor groups like the California Teamsters, are petitioning California Attorney General Rob Bonta to stop OpenAI from becoming a for-profit entity, The Los Angeles Times reports. OpenAI announced plans to transition to a public-benefit corporation in 2024, and reportedly has two years to pull it off or risk a large portion of the money its raised become debt. The group's primary concerns are that OpenAI "failed to protect its charitable assets" and is actively "subverting its charitable mission to advance safe artificial intelligence." OpenAI started as a nonprofit research organization studying AI, but transitioned to a for-profit company that's overseen and run by a nonprofit in 2019. That structure is legally allowed in the state of California, but the group's petition claims that OpenAI's decision to pursue a new structure is driven by a desire not to further its mission, but to provide "AI's benefits -- the potential for untold profits and control over what may become powerful world-altering technologies -- to a handful of corporate investors and high-level employees."

Anthropic is joining the ranks of OpenAI in offering a more expensive tier of its flagship chatbot. On Wednesday, the company announced Max Plan. Starting today, you can either pay 100 or 200 per month to use Claude up to 5x or 20x more than you can with Anthropic's existing Pro plan. The company told Engadget it's introducing the Max tier in response to the popularity of Claude 3.7 Sonnet. The new hybrid reasoning model, which excels at coding tasks, has been so popular with users, many are asking to use it as much as they want.