ViStoryBench is a comprehensive evaluation benchmark for story visualization frameworks, featuring diverse datasets and metrics to assess model performance across narrative and visual dimensions.

This paper talks about ViStoryBench, a new set of tests designed to check how well AI models can turn written stories into pictures by using a variety of story types and ways to measure their performance.

ViStoryBench: Comprehensive Benchmark Suite for Story Visualization

LLMSynthor enhances LLMs for efficient and statistically accurate data synthesis through distributional feedback and proposal sampling.

This paper talks about LLMSynthor, a new method that helps large language models create fake data that looks and behaves like real data, making the process faster and more accurate.

Large Language Models for Data Synthesis

v1 enhances Multimodal Large Language Models by enabling selective and dynamic visual region retrieval during inference, improving performance on multimodal reasoning tasks.

This paper talks about a new approach called v1 that helps AI models do a better job at understanding and reasoning about both images and text by letting them go back and look at specific parts of an image more than once while answering questions.

Don't Look Only Once: Towards Multimodal Interactive Reasoning with
  Selective Visual Revisitation

DINO-R1 incorporates reinforcement learning to enhance visual in-context reasoning capabilities in vision foundation models, achieving better performance than supervised fine-tuning across various visual prompting scenarios.

This paper talks about DINO-R1, a new way to train vision foundation models so they get better at reasoning about what they see in images by using reinforcement learning instead of just regular training.

DINO-R1: Incentivizing Reasoning Capability in Vision Foundation Models

AlphaOne dynamically modulates reasoning in large models by introducing $\alpha$ moment and Bernoulli process for slow thinking, improving efficiency and capability across diverse domains.

This paper talks about AlphaOne, a new approach that lets AI models switch between thinking quickly or slowly depending on the situation, which helps them solve problems more efficiently and effectively.

AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

Open CaptchaWorld benchmark evaluates MLLM-powered agents on diverse CAPTCHA puzzles, revealing significant performance gaps compared to humans.

This paper talks about Open CaptchaWorld, a new online platform that tests how well AI models can solve different types of CAPTCHA puzzles, which are the challenges websites use to tell humans and bots apart.

Open CaptchaWorld: A Comprehensive Web-based Platform for Testing and
  Benchmarking Multimodal LLM Agents

HARDTESTGEN creates a large, high-quality competitive programming dataset to enhance the precision and recall of verifiers in evaluating LLM-generated code.

This paper talks about HARDTESTGEN, a tool that makes lots of tough and well-designed test cases for coding problems, which helps check if the code written by large language models is actually correct and reliable.

HardTests: Synthesizing High-Quality Test Cases for LLM Coding

EXP-Bench evaluates AI agents' end-to-end research experiment capabilities through curated tasks from top AI papers, highlighting current limitations.

This paper talks about EXP-Bench, a new test that checks if AI agents can actually carry out full research experiments on their own by using tasks taken from some of the best AI research papers.

EXP-Bench: Can AI Conduct AI Research Experiments?

A coordinated diffusion noise optimization framework improves whole-body manipulation of articulated objects by leveraging specialized diffusion models for body and hand motions and a unified basis point set representation for precise hand-object interaction.

This paper talks about CoDA, a new method that helps robots or AI systems move and control objects with multiple moving parts, like doors or tools, by making their body and hand movements much more coordinated and precise.

CoDA: Coordinated Diffusion Noise Optimization for Whole-Body
  Manipulation of Articulated Objects

A framework for post-training multimodal large language models using reinforcement learning with verifiable rewards introduces a data mixture strategy to enhance general reasoning abilities and benchmark performance.

This paper talks about MoDoMoDo, a new system that helps AI models learn to reason better across different types of data, like text, images, and more, by using reinforcement learning and a special mix of training data.

MoDoMoDo: Multi-Domain Data Mixtures for Multimodal LLM Reinforcement
  Learning

SpookyBench is a benchmark for temporal pattern recognition in videos that highlights the limitations of vision-language models in processing noise-like frames without spatial information.

This paper talks about SpookyBench, a new test designed to see how well AI models that understand both video and language can recognize patterns over time in videos, especially when the frames look random and don't have clear shapes or objects.

Time Blindness: Why Video-Language Models Can't See What Humans Can?

Prolonged reinforcement learning training (ProRL) uncovers novel reasoning strategies in language models, outperforming base models and suggesting meaningful expansion of reasoning capabilities.

This paper talks about ProRL, a method that trains language models for a longer time using reinforcement learning, which helps these models come up with new and better ways to solve problems and reason through tough questions.

AI Research Made Simple

ViStoryBench: Comprehensive Benchmark Suite for Story Visualization

Large Language Models for Data Synthesis

Don't Look Only Once: Towards Multimodal Interactive Reasoning with Selective Visual Revisitation

DINO-R1: Incentivizing Reasoning Capability in Vision Foundation Models

AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

Open CaptchaWorld: A Comprehensive Web-based Platform for Testing and Benchmarking Multimodal LLM Agents

HardTests: Synthesizing High-Quality Test Cases for LLM Coding

EXP-Bench: Can AI Conduct AI Research Experiments?

CoDA: Coordinated Diffusion Noise Optimization for Whole-Body Manipulation of Articulated Objects

MoDoMoDo: Multi-Domain Data Mixtures for Multimodal LLM Reinforcement Learning

Time Blindness: Why Video-Language Models Can't See What Humans Can?

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models