Footstep recognition identified as new biometric identifier, combining sound & vibration patterns from walking, similar to fingerprints.

New AI model, ASSNet, achieves record accuracy in detecting tiny tumors & organs in medical scans using Vision Transformer-based architecture with adaptive attention mechanisms.

Flux-1.1-Pro is a powerful text-to-image AI model by Black-Forest-Labs, offering fast generation & improved image quality, prompt adherence & output diversity.

Large language models exhibit social desirability bias in survey answers, mirroring human behavior and aligning with social norms.

AI Gets Smarter: New Method Helps Systems Learn 40% Faster Across Different Environments, combining transfer learning with contextual reinforcement learning for improved sample efficiency and performance.

New GUI grounding approach boosts accuracy by 15% through iterative narrowing and multiple refinement steps, enhancing desktop automation and accessibility.

Alan Health creates AI "Mo" for patient chats, built with large language models & custom medical knowledge, serving 200k+ users in France.

LLaMA-Berry model solves math Olympiad problems like human experts using pairwise optimization, demonstrating strong performance on challenging tasks.

Brain-inspired pruning method cuts Neural Networks by 90% without losing accuracy, using criticality theory to identify important neurons and adaptively prune less useful ones.

Wavelet-Based AI Model outperforms top performers in image generation, eliminating need for Vector Quantization. Novel autoregressive model uses wavelets to capture multi-scale dependencies efficiently.

LLMs show promise in PBE tasks but struggle with new problem types, fine-tuning improves performance but out-of-distribution generalization remains a challenge.

Bio-Inspired Neural Networks cut 3D scene rendering costs by 95% while maintaining quality with Spiking NeRF, a combo of neural radiance fields & bio-inspired spiking neural networks.

New AI method StableV2V for shape-consistent video editing breaks down editing into sequential steps, aligns motion patterns with user prompts and outperforms existing methods in consistency and efficiency.

GazeGen uses gaze-driven user interaction for visual content generation, allowing users to guide image creation with their eyes.

AI models show different paths to abstract reasoning: Function vs Direct Prediction. Two approaches explored: inferring latent functions or directly predicting new test outputs using neural networks on ARC dataset.

Research examines continuous-time models of adaptive optimization algorithms, focusing on AdaGrad, RMSProp & Adam optimizers, proving convergence properties.

Test-time training boosts AI model's abstract reasoning by 30% on ARC benchmark, study shows.

Research paper critiques skepticism around AI in chip design, addressing reproduction errors and methodological flaws.

New AI model LLaVA-o1 boosts accuracy by 15% on visual tasks with step-by-step reasoning, mirroring human detective work.

KnowAda bridges "visual gap" with knowledge-adapted captions, boosting performance on complex visual reasoning tasks.

Qwen-7B-Chat is a 7 billion param AI model, pre-trained on web texts & code. It generates responses to text prompts, with capabilities in natural language processing tasks.

Large language models (LLMs) can self-improve in long-context reasoning through proper prompting strategies, enhancing their ability to understand and generate human-like text.

Quantum computer makers urged to stop overstating performance, misleading public with "fool the masses" tactics, instead adopt transparent reporting standards.

LLMs in robots vulnerable to "jailbreaking" attacks, researchers introduce RoboPAIR algorithm to elicit harmful physical actions.

GPTree combines LLMs & decision trees for explainable decision-making, generating natural language explanations for predictions on founder success dataset.

Data Prep Kit (DPK) simplifies & scales data prep for LLMs, allowing users to prepare data locally or on a cluster with thousands of CPU cores.

RedCode benchmark evaluates AI code agent safety. It tests recognition & handling of unsafe code, as well as generation of harmful code when given prompts.

Discovering anomalies in complex networks with UniGAD: A Multi-Level Graph Approach introduces a new method for detecting anomalous nodes/edges in graph-structured data using spectral subgraph sampling.

Large language models require significant resources, but LLM-Neo distills knowledge into smaller models efficiently.

Video generation aims to model authentic & customized motion across frames. Diffusion-based studies lack interpretability & transparency in encoding cross-frame motion info.

CDLGNs combine deep learning & logical operations for interpretable AI solutions. They can learn & represent logical functions, solving complex tasks with clarity & flexibility.

New byte encoding scheme, MYTE, boosts multilingual AI fairness & performance by leveraging morphological info for more effective character encoding.

LLMs used for hyperparameter optimization efficiently navigate search space & identify optimal configurations in machine learning models.

Revolutionary AI Qwen2.5-Coder boosts coding tasks with improved code gen, understanding & debugging capabilities.

Mathematical framework proposes Riemannian geometry for understanding intelligence & consciousness, linking neural reps to thought processes.

Multimodal models require massive data & compute. FuseMix uses pre-trained encoders for efficient multimodal alignment on a single GPU, making it accessible for practical use cases.

Stable-Diffusion-V1-4: AI model for generating images. Simplified guide by Compvis, subscribe to AImodels.fyi newsletter or follow on Twitter for more guides.

API-protected LLMs leak proprietary details through logits, a "back door" that reveals model training data & objective function. Researchers find API calls can extract full logit vector, compromising IP of LLM providers.

ADOPT algorithm outperforms Adam in certain cases by converging at optimal rate regardless of β₂ value, addressing a key limitation of Adam.

Agent K v1.0 automates data science tasks with self-learning, achieving 92.5% success rate & rivaling expert-level human competitors on Kaggle.

Small language models (SLMs) play a crucial role in natural language processing alongside large language models (LLMs). They're compact, efficient & can collaborate with LLMs to leverage strengths of both model types.

Language model prompting is Turing complete, meaning it can perform any possible computation. Researchers demonstrated how to design prompts that simulate a given Turing machine, showing the unbounded computational power of language models.

Expert human forecasters outperformed top-performing LLM in statistically significant way (p-value = 0.01) on ForecastBench, a new dynamic benchmark for evaluating forecasting capabilities of ML systems.

Pancomputational enactivism grounds consciousness in fundamental computational processes, making it a universal feature of the physical world, not limited to brains or biological systems.

sd-inpaint model fills masked areas of images using Stable Diffusion, generating high-quality inpainted images with seamless blending. Use it to remove unwanted objects, complete partially obscured images, or create new art within existing images.

Image captioning just got a boost with the ImageInWords dataset, containing 2.5M image-description pairs with hyper-detailed descriptions of images. This could aid tasks like accessibility & visual question answering.

BitsFusion quantizes diffusion model weights to 1.99 bits avg, maintaining high performance & efficiency. Outperforms other methods on image generation & text-to-image tasks.

Brain-like inference uses entropy-minimizing algorithm inspired by variational inference & neuroscience. New objective function & algorithm proposed to efficiently process info & make inferences.

Generator matching unifies generative models like diffusion & flow matching using Markov processes, enabling superpositions & multimodal models for complex data distributions.

Researchers propose "conditional hallucinations" method for image compression, generating missing details to maintain visual quality & achieve better compression ratios.

Chain-of-Thought reasoning improves performance on complex tasks but can reduce it when humans rely on intuition over analysis, leading to "overthinking" and suboptimal choices in some cases.

Replicating O1 model: Researchers shift from "shortcut learning" to "journey learning", gaining valuable insights & advancing AI research. Chronological overview of steps taken & key findings shared in progress report.

LLMs like GPT-4 excel as text classifiers, matching traditional models in various domains & even exceeding performance in some cases. They also show promise for few-shot learning & fine-tuning, making them a powerful tool for smart expert systems.

Researchers propose Infinite Context LLMs to mimic human episodic memory, enabling models to recall past experiences and adapt to new situations.

Breaking memory limits in contrastive learning: researchers introduce "Near Infinite Batch Size Scaling" (NIBS) method, achieving significant performance gains on various benchmarks with much larger effective batch sizes.

LLMs can perform tasks like writing essays & answering questions but still have limitations compared to humans in certain domains, raising concerns about job market impacts & human-AI collaboration.

NVLM: Frontier-Class Multimodal LLMs combine language, vision & more into seamless versatile AI models. Enables new apps that tightly integrate different data types, but poses significant computational & safety challenges.

Language models can learn about themselves through introspection, developing self-knowledge of strengths, weaknesses & biases. This ability could enhance reliability & transparency in AI systems.

Diffusion transformer models can be compressed 4.5x with new quantization technique PTQ4DiT while preserving image quality. This makes powerful AI-driven image generation accessible on resource-constrained devices like smartphones.

Meissonic model breaks through in text-to-image synthesis, matching state-of-the-art diffusion models with non-autoregressive MIM approach & high-quality training data.

16-bit precision in ML models can match 32-bit accuracy & boost speed, especially valuable for practitioners with limited hardware resources due to its widespread availability across GPUs.

Large Language Models (LLMs) can perform multiple tasks simultaneously during a single inference call, a capability called "task superposition", defying the assumption that LLMs learn one task at a time.

New AI-powered image editor lets users control images with natural language prompts, offering versatile & disentangled control over object attributes, scene composition & style.

New approach to evaluate AI's semantic significance: 'Semantic Importance Betting' (SIB) task where humans bet on text importance. SIB yields insights missed by standard metrics, guiding future model development.

Quantum autoencoders outperform classical deep learning models in anomaly detection by 60-230 times with fewer parameters & iterations, opening doors to solving complex time series data problems.

LLMs can be viewed as Markov chains, predicting next word based only on current state, without considering full history. In-context learning (ICL) allows LLMs to adapt predictions based on provided context.

Selective Attention boosts Transformer performance by 0.5-2.0 BLEU points on tasks like machine translation & question answering. It selectively attends to a subset of input elements, improving efficiency & accuracy.

New code completion feature for IntelliJ IDEs suggests entire lines of code, working locally & securely, with focus on speed & efficiency.

GPUDrive simulates realistic driving at 1M FPS with data-driven multi-agent system, enabling scalable & detailed urban scenarios with thousands of vehicles & pedestrians.

MANTRA dataset: 1000+ triangulation meshes representing manifold surfaces for topological data analysis, geometric deep learning & manifold learning research.

Neural theorem prover miniCTX boosts performance by 30% with long-context understanding, outperforming prior models on standard benchmarks.

WatChat system helps explain complex code by debugging users' mental models through natural language interaction, improving comprehension in user study.

CMA-ES Optimizer improved with adaptive learning rate for faster convergence & better solutions on black-box optimization problems.

Researchers propose TPI-LLM to run large language models on low-resource edge devices, splitting the model across multiple devices & reducing memory footprint, achieving comparable performance with lower resource requirements.

Top models tackle billion-scale nearest neighbor search at NeurIPS'23, showcasing novel neural networks & optimization techniques for efficient ANN search with applications in image retrieval & product recommendation.

TypeFly system lets users control drones with plain English instructions using a large language model. It translates natural language into drone actions, making it easier for non-technical people to fly drones.

o1 AI model shows promise in medical tasks like diagnosis & treatment recommendations but has limitations in understanding context & patient interactions. Further research needed to develop "AI doctor" capable of replacing human physicians.

Researchers propose a "cross-chirality" palmprint verification system matching left to right palms & vice versa, expanding real-world applications of palmprint biometrics.

Transformers get thought-provoking with Chain of Thought reasoning: models generate step-by-step explanations to solve complex tasks like math problems & multi-hop question answering.

MINT-1T: Open-Source Multimodal Dataset with 1 trillion tokens, enabling more capable AI models. Researchers can train robust multimodal models with diverse text, images & other modalities.

Iterative Thought Refiner enhances LLM responses via dynamic adaptive reasoning, leveraging human engagement to refine answers.

Researchers break reCAPTCHAv2 with machine learning, combining image classification & segmentation to solve challenges with high accuracy, raising questions about long-term viability of "proof-of-personhood" systems.

Large language models (LLMs) improved with Critical Planning Step Learning (CPL) & Step-level Advantage Preference Optimization (Step-APO), boosting general reasoning capabilities across various domains.

Large Language Models improved by 112.93% with REAP method, enhancing problem-solving & output clarity.

Large language models can now improve themselves without explicit human guidance thanks to the "ImPlicit Self-ImprovemenT" (PIT) framework, which learns improvement goals from human preference data.

Large language models can unlock novel scientific research ideas by generating innovative hypotheses & accelerating scientific progress in various fields.

New AdEMAMix optimizer blends existing techniques for better performance, faster convergence, and stable training. It combines Adam & AMSGrad strengths to achieve improved results in various benchmarks.

Large language models like GPT-3, BERT & RoBERTa excel in log parsing tasks, outperforming traditional methods with high accuracy & efficiency.

Fine-tuned Vision Transformer (ViT) model by Falcons-Ai classifies images as "normal" or "not safe for work" (NSFW). Developed from pre-trained ViT architecture, it accurately distinguishes between safe & explicit visual content.

Model merging combines LLMs & MLLMs into a single powerful model, enabling better performance on various tasks & improving accessibility in low-resource settings.

Wash trading in Ethereum NFT market: $3.4B artificial volume, 5.66% of collections affected, exploiting reward systems more profitable than price inflation.

RLAIF outperforms RLHF in aligning LLMs with human preferences, offering a scalable solution to costly human feedback.

Paper proposes compressing data into low-dimensional Gaussian mixture using CRATE models, achieving competitive results with transformer-based models.

Smaller LLMs outperform larger models in reasoning tasks with "compute-optimal sampling" training approach, reducing model size & compute requirements while maintaining performance.

LLMs generate functional code but struggle with security & testability issues, researchers find in new study on Assessing LLM Code Generation: Quality, Security and Testability Analysis.

Large language models may overstate safety & reliability. Researchers propose training LLMs to better recognize & communicate limitations & uncertainties, improving transparency & calibration.

AI hype leads to unrealistic expectations & public disappointment. Researchers must provide balanced, nuanced assessments of AI capabilities to maintain trust & prevent harm.

Diffusion models can revolutionize real-time game engines, enabling dynamic & immersive virtual worlds that respond to user inputs in real-time, potentially changing the video game industry forever.

Researchers use machine learning to develop AI that mimics pro Counter-Strike players' movement patterns & strategies, with potential apps in game bot dev, virtual training & player control design.

RAG: Enhancing Large Language Models with External Knowledge for Informative Text Generation combines LLMs with external knowledge sources to generate more informative & coherent text.