Neural intel Pod

Recent progress in video generation still struggles with issues like motion instability and prompt alignment. To address this, the study explores incorporating human preferences into advanced flow-based video generation models. The authors introduce a large, new dataset of human-annotated video preferences across visual quality, motion quality, and text alignment. They also develop a multi-dimensi...

This research paper introduces  AnimeGamer , a novel method for creating  infinite anime life simulation games . Unlike previous work limited to static images or finite gameplay, AnimeGamer uses a  multimodal large language model  to process user instructions and predict subsequent game states. These states include  dynamic animation shots and updated character attributes , aiming for a more immer...

The provided research introduces NoProp, a novel method for training neural networks that deviates from traditional backpropagation by eliminating both forward and backward passes. Instead, it draws inspiration from diffusion models, where each layer learns independently to denoise a noisy version of the target. This approach trains networks without hierarchical representation learning in the conv...

The provided paper introduces  ACPBench Hard , a new benchmark designed to evaluate the reasoning capabilities of large language models for automated planning. This benchmark extends the original ACPBench by featuring  open-ended, generative versions of planning-related questions  across various tasks, mirroring the challenges faced by symbolic planners. The authors tested several large language a...

From Amazon's blog

Dynamic 4D Modeling from Casual Video

The provided research paper introduces KDTalker, a novel method for generating realistic audio-driven talking portraits by combining implicit 3D keypoints with a spatiotemporal diffusion model. This framework addresses limitations in existing techniques by achieving high lip synchronization accuracy and diverse head poses while maintaining computational efficiency. KDTalker leverages unsupervised...

AI2 has announced OLMo 2 , a new family of fully open 7B and 13B language models demonstrating  performance on par with or exceeding similarly sized open models , even rivaling open-weight models like Llama 3.1. This release includes not only model weights but also  data, code, and evaluation frameworks , emphasizing truly open science. Key advancements in OLMo 2 involve  improved training stabili...

Stable 3D Shape Correspondence via Registration

The provided text introduces  ProjectEval , a new benchmark for automatically evaluating the project-level code generation capabilities of programming agents by simulating user interactions.  ProjectEval  aims to address the limitations of existing benchmarks, such as a lack of automated user-centric evaluation and result explainability. The benchmark includes diverse real-world tasks with varying...

This research introduces a novel framework using  elicitation and execution policies  to improve how  embodied AI agents  express their confidence in complex, open-ended environments like Minecraft. The authors demonstrate that  structured reasoning techniques , such as Chain-of-Thought and Plan & Solve, significantly enhance an agent's ability to  calibrate its confidence and predict fail...

The provided research introduces  VLM-Attention , a novel StarCraft II environment designed to better reflect human perception and decision-making by incorporating RGB visuals and natural language. This framework utilizes vision-language models with specialized mechanisms for unit targeting, knowledge retrieval for tactical decisions, and dynamic role assignment for coordinated multi-agent behavio...

The provided research paper introduces a novel attack method,  M-Attack , designed to effectively fool sophisticated commercial large vision-language models like GPT-4.5 and Gemini. The paper highlights the  limitations of existing attack strategies  which often produce uniform and semantically vague perturbations, failing against these robust models.  M-Attack overcomes these issues  by focusing...

This paper introduces a  deep learning-based method  for the  generalized inverse design  of complex  multi-port radio-frequency and sub-terahertz electromagnetic structures  and integrated circuits. This innovative approach uses  deep neural networks  to accurately predict the properties of arbitrary-shaped designs, significantly  speeding up the synthesis process  compared to traditional methods...

Cool twitter find

R1 with vision?

The provided files detail  OWL , an innovative framework built upon CAMEL-AI, designed for  multi-agent collaboration  to automate real-world tasks. The README.md offers a comprehensive overview of OWL's  features , including toolkits for web interaction, document processing, and code execution, along with  installation instructions  using various methods like uv, venv, conda, and Docker. The ...

That

Unsloth's documentation provides a comprehensive guide to fine-tuning large language models , particularly for beginners.  It outlines the benefits of fine-tuning , such as improved domain knowledge and task-specific optimization, and compares it to methods like RAG.  The guide details essential steps , including choosing the right model and fine-tuning method like LoRA or QLoRA, preparing dat...

Updates from the PyTorch blog

This research introduces a novel, deep-learning-enabled approach for designing complex multi-port radio-frequency and sub-terahertz circuits and electromagnetic structures.  The methodology overcomes limitations of traditional design methods by using deep learning models to achieve rapid synthesis of arbitrary-shaped structures with desired radiative and scattering properties.  The key innovation...

This research addresses the challenge of improving monocular depth estimation (MDE) using unlabeled data through a novel distillation framework.   The core innovation is Cross-Context Distillation, which combines local and global depth cues to enhance pseudo-label quality and model accuracy.   A multi-teacher distillation approach further leverages complementary strengths of different depth estima...

The research introduces MHA2MLA, a novel fine-tuning framework designed to adapt existing MHA-based language models to the more efficient MLA architecture.   MLA achieves economical inference by compressing the key-value (KV) cache.   MHA2MLA employs partial RoPE and low-rank approximation techniques to minimize performance degradation during the adaptation.   Experiments demonstrate that MHA2MLA,...

This paper introduces SWE-RL, a reinforcement learning (RL) method to improve large language models (LLMs) for software engineering tasks using software evolution data and rule-based rewards.  The approach trains LLMs to autonomously learn from open-source software's lifecycle, including code snapshots, changes, and events.  The resulting model, Llama3-SWE-RL-70B, achieves state-of-the-art per...

Amazon Science highlights research and career opportunities across various fields, with a focus on quantum computing.  It showcases a specific project where SAT solving is used to optimize quantum circuit mapping, leading to significant speed improvements. The method aims to reduce the number of swap gates in quantum circuits, a crucial factor in quantum computation efficiency.  The company is act...