AI Breakdown

In this episode, we discuss MEGA-Bench: Scaling Multimodal Evaluation to over 500 Real-World Tasks by Jiacheng Chen, Tianhao Liang, Sherman Siu, Zhengqing Wang, Kai Wang, Yubo Wang, Yuansheng Ni, Wang Zhu, Ziyan Jiang, Bohan Lyu, Dongfu Jiang, Xuan He, Yuan Liu, Hexiang Hu, Xiang Yue, Wenhu Chen. The paper introduces MEGA-BENCH, a comprehensive evaluation suite featuring over 500 real-world multimodal tasks to address diverse daily user needs. It includes more than 8,000 samples curated by 16 expert annotators, utilizing a variety of output formats such as numbers, phrases, and code instead of standard multiple-choice questions. MEGA-BENCH aims to provide high-quality, diverse data for cost-effective and accurate model evaluation across a wide range of multimodal tasks.

In this episode, we discuss TrajectoryCrafter: Redirecting Camera Trajectory for Monocular Videos via Diffusion Models by Mark YU, Wenbo Hu, Jinbo Xing, Ying Shan. TrajectoryCrafter is a new method that precisely redirects camera paths in monocular videos by separating view changes from content generation. It uses a dual-stream conditional video diffusion model that combines point cloud renders with source videos to ensure accurate views and coherent 4D content. By training on a hybrid dataset of monocular and multi-view videos with a double-reprojection strategy, TrajectoryCrafter achieves robust performance across diverse scenes.

In this episode, we discuss PlanGEN: A Multi-Agent Framework for Generating Planning and Reasoning Trajectories for Complex Problem Solving by Mihir Parmar, Xin Liu, Palash Goyal, Yanfei Chen, Long Le, Swaroop Mishra, Hossein Mobahi, Jindong Gu, Zifeng Wang, Hootan Nakhost, Chitta Baral, Chen-Yu Lee, Tomas Pfister, Hamid Palangi. The paper introduces **PlanGEN**, a versatile agent framework designed to tackle complex planning problems by incorporating constraint, verification, and selection agents. PlanGEN enhances existing inference-time algorithms through constraint-guided iterative verification and dynamically selects the optimal algorithm based on the complexity of each instance. Experimental results show that PlanGEN significantly outperforms leading baselines across multiple benchmarks, achieving state-of-the-art performance by effectively improving verification processes and adaptive algorithm selection.

In this episode, we discuss VideoGrain: Modulating Space-Time Attention for Multi-grained Video Editing by Xiangpeng Yang, Linchao Zhu, Hehe Fan, Yi Yang. The paper introduces VideoGrain, a zero-shot method that enhances multi-grained video editing by modulating space-time attention mechanisms for class-, instance-, and part-level modifications. It addresses challenges like semantic misalignment and feature coupling by improving text-to-region control and optimizing feature separation within diffusion models. Extensive experiments demonstrate that VideoGrain achieves state-of-the-art performance in real-world video editing scenarios.

In this episode, we discuss ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Models by Jonathan Roberts, Mohammad Reza Taesiri, Ansh Sharma, Akash Gupta, Samuel Roberts, Ioana Croitoru, Simion-Vlad Bogolin, Jialu Tang, Florian Langer, Vyas Raina, Vatsal Raina, Hanyi Xiong, Vishaal Udandarao, Jingyi Lu, Shiyang Chen, Sam Purkis, Tianshuo Yan, Wenye Lin, Gyungin Shin, Qiaochu Yang, Anh Totti Nguyen, Kai Han, Samuel Albanie. The paper reveals that Large Multimodal Models (LMMs) have significant difficulties with image interpretation and spatial reasoning, often underperforming compared to young children or animals. To address this gap, the authors introduce ZeroBench, a challenging visual reasoning benchmark comprising 100 carefully designed questions and 334 subquestions that current LMMs cannot solve. Evaluation of 20 models resulted in a 0% score on ZeroBench, and the benchmark is publicly released to stimulate advancements in visual understanding.

In this episode, we discuss Teaching Language Models to Critique via Reinforcement Learning by Zhihui Xie, Jie chen, Liyu Chen, Weichao Mao, Jingjing Xu, Lingpeng Kong. The paper presents CTRL, a framework that uses reinforcement learning to train critic models which provide feedback for improving code generated by large language models without needing human input. These trained critics significantly increase code pass rates and reduce errors across different generator models. Additionally, the critics serve as effective reward models, allowing iterative refinements that lead to over 106% improvement on challenging code generation benchmarks.

In this episode, we discuss PANDAS: Improving Many-shot Jailbreaking via Positive Affirmation, Negative Demonstration, and Adaptive Sampling by Avery Ma, Yangchen Pan, Amir-massoud Farahmand. The paper introduces PANDAS, a hybrid technique that enhances many-shot jailbreaking by altering fabricated dialogues with positive affirmations, negative demonstrations, and optimized adaptive sampling tailored to specific prompts. Experimental results on AdvBench and HarmBench using advanced large language models show that PANDAS significantly outperforms existing baseline methods in scenarios involving long input contexts. Additionally, an attention analysis highlights how PANDAS exploits long-context vulnerabilities, providing deeper insights into the mechanics of many-shot jailbreaking.

In this episode, we discuss VidCRAFT3: Camera, Object, and Lighting Control for Image-to-Video Generation by Sixiao Zheng, Zimian Peng, Yanpeng Zhou, Yi Zhu, Hang Xu, Xiangru Huang, Yanwei Fu. The paper presents VidCRAFT3, a new framework for image-to-video generation that allows simultaneous control over camera motion, object movement, and lighting direction. It addresses previous limitations by introducing the Spatial Triple-Attention Transformer, which effectively decouples and integrates lighting, text, and image inputs. This innovative approach enhances the precision and versatility of controlling multiple visual elements in generated videos.

In this episode, we discuss Heuristically Adaptive Diffusion-Model Evolutionary Strategy by Benedikt Hartl, Yanbo Zhang, Hananel Hazan, Michael Levin. The paper explores the connection between diffusion models and evolutionary algorithms, highlighting that both generate high-quality samples through iterative refinement of random initial states. By integrating deep learning-based diffusion models into evolutionary processes, the authors enhance convergence efficiency and maintain diversity by leveraging improved memory and refined sample generation. This framework advances evolutionary optimization by providing greater flexibility, precision, and control, representing a significant shift in heuristic and algorithmic approaches.

In this episode, we discuss Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach by Jonas Geiping, Sean McLeish, Neel Jain, John Kirchenbauer, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein. The paper presents a new language model architecture that enhances test-time computation by iteratively reasoning in latent space using a recurrent block, allowing flexible depth during inference. Unlike chain-of-thought approaches, it doesn't require specialized training data, works with small context windows, and can handle complex reasoning not easily expressed in words. A 3.5 billion parameter model was scaled to 800 billion tokens, demonstrating significant performance improvements on reasoning benchmarks with computation loads up to 50 billion parameters. Huggingface: https://huggingface.co/papers/2502.05171 Github: https://github.com/seal-rg/recurrent-pretraining