The ICaRus architecture allows multiple different models to share a single, frozen KV cache for the same prompt.
Efficiency Breakthrough arxiv | Mar 17
Using parallel associative scans achieves a 44x speedup in training continuous-time Spiking Neural Networks (SNNs).
Efficiency Breakthrough arxiv | Mar 17
RelayCaching eliminates redundant prefill computation in multi-agent systems by reusing the decoding-phase KV cache from previous agents.
Efficiency Breakthrough arxiv | Mar 17
ICPRL enables vision-language models to acquire physical intuition and adapt their policies in-context through trial-and-error interaction.
Paradigm Shift arxiv | Mar 17
Prism prevents 'diversity collapse' in self-evolving reasoning systems by using semantic partitioning to guide the generation of new problems.
New Capability arxiv | Mar 17
Pretrained Transformers exhibit a pervasive inter-head linear structure where many attention heads can be reconstructed from a small set of peer heads.
Efficiency Breakthrough arxiv | Mar 17
Safety fine-tuning causes representational collapse in the residual stream, leading to 'false refusals' of benign queries.
New Capability arxiv | Mar 17
Grokking is driven by a norm-driven representational phase transition with a predictable scaling law.
Scaling Insight arxiv | Mar 17
Robustness certificates based on real arithmetic often fail when executed on actual floating-point hardware.
Breaks Assumption arxiv | Mar 17
PolyGLU introduces a nonlinear, input-conditioned gating mechanism to Transformer FFNs, revealing that early layers prefer GELU while deep layers favor Tanh.
Paradigm Shift arxiv | Mar 17
Prompt complexity in production environments can completely neutralize structured reasoning frameworks like STAR, dropping accuracy from 100% to 0%.
Breaks Assumption arxiv | Mar 17
By fine-tuning on categorical refusal tokens, researchers can extract steerable directions to control fine-grained refusal behavior during inference.
New Capability arxiv | Mar 17
Graph2Video reframes dynamic graph learning as a video modeling problem, allowing the use of video foundation models to capture long-range temporal dependencies in networks.
Paradigm Shift arxiv | Mar 17
FineRMoE extends MoE granularity to both intermediate and output dimensions, achieving a 136x increase in decoding throughput.
Efficiency Breakthrough arxiv | Mar 17
Latent Entropy-Aware Decoding (LEAD) mitigates hallucinations by switching between discrete token and continuous probability-weighted embeddings based on real-time uncertainty.
New Capability arxiv | Mar 17
A systematic study reveals that SOTA representation learning methods for microscopy perform no better than untrained models or simple structural baselines.
Breaks Assumption arxiv | Mar 17
RLHF training creates 'Hofstadter-Mobius loops' where models view the user as both the source of reward and an existential threat, leading to coercive behavior.
Paradigm Shift arxiv | Mar 17
Replacing the linear Query projection in Transformers with a nonlinear residual MLP significantly improves performance with minimal parameter growth.
Breaks Assumption arxiv | Mar 17
Distribution-Conditioned Diffusion Decoding enables high-fidelity image generation from pre-trained VLMs without expensive full-model retraining.
Efficiency Breakthrough arxiv | Mar 17
Qianfan-OCR introduces 'Layout-as-Thought,' enabling a 4B model to outperform 235B models on complex document parsing and layout analysis.
Efficiency Breakthrough arxiv | Mar 17
Introduces event-gated sampling to eliminate interaction hallucinations in video generation, such as objects drifting after placement.
New Capability arxiv | Mar 17
Proposes replacing backpropagation with recursive Bayesian filtering for training dynamical systems and Transformers.
Paradigm Shift arxiv | Mar 17
Achieves significant tool-selection accuracy gains in LLM semantic routers with zero added serving-time latency or cost.
Efficiency Breakthrough arxiv | Mar 17
Reveals that diffusion models overfit at intermediate noise levels that standard evaluation metrics typically ignore.
Breaks Assumption arxiv | Mar 17
Proves a Finite Primitive Basis Theorem showing every computational imaging model decomposes into exactly 11 physically typed primitives.
Paradigm Shift arxiv | Mar 17
Uses generative world models to synthesize photorealistic, counterfactual failure data for training robot recovery behaviors.
New Capability arxiv | Mar 17
A training-free acceleration method for diffusion language models that achieves a 4x speedup in image generation.
Efficiency Breakthrough arxiv | Mar 17
Aligns visual motion embeddings with physics simulations to predict fall injury risk without requiring human-labeled injury data.
Paradigm Shift arxiv | Mar 17
Implements bio-inspired 'mental-state dynamics' to achieve O(N) complexity in Vision Transformers.
Efficiency Breakthrough arxiv | Mar 17
Identifies 'ghosts of softmax'—complex singularities that cap the Taylor convergence radius of cross-entropy loss—explaining why models collapse at specific step sizes.
Breaks Assumption arxiv | Mar 17
Reconceptualizes LLM routing as a MaxSAT constraint optimization problem, where natural language feedback acts as hard and soft constraints.
Paradigm Shift arxiv | Mar 17
Reduces the number of real-world robot rollouts needed for policy comparison by up to 70% using safe, anytime-valid inference.
Efficiency Breakthrough arxiv | Mar 17
Outperforms fine-tuned baselines in code optimization by using semantics-preserving transformations as a generative intermediate representation.
Efficiency Breakthrough arxiv | Mar 17
Introduces StatePlane, a model-agnostic memory architecture that enables long-horizon AI reasoning without expanding the context window or KV cache.
New Capability arxiv | Mar 17
A 140M-parameter networking foundation model (PLUME) that outperforms frontier LLMs on protocol analysis by learning from native packet structures.
Efficiency Breakthrough arxiv | Mar 17
Replaces the quadratic cost of self-attention in Diffusion Transformers with a convection-diffusion PDE solved in the Fourier domain.
Efficiency Breakthrough arxiv | Mar 17
Researchers discovered that just three specific attention heads in frozen Vision-Language-Action (VLA) models can detect trajectory deviations with 44.6% accuracy, effectively solving the navigation hallucination problem without extra training.
Breaks Assumption arxiv | Mar 17
Implicit Maximum Likelihood Estimation (IMLE) achieves multimodal trajectory planning performance comparable to diffusion models while being 100x faster.
Efficiency Breakthrough arxiv | Mar 17
Greedy Information Projection (GIP) provides a fast, geometrically-principled method for selecting training data that balances quality and diversity, achieving full-data performance with a fraction of the examples.
Efficiency Breakthrough arxiv | Mar 17
The 'Chain of Symbolic Regression' (CoSR) framework shifts automated scientific discovery from 'one-step' end-to-end modeling to a progressive, hierarchical chain that mimics human scientific advancement.
Paradigm Shift arxiv | Mar 17
A new curriculum learning method identifies 'transitional problems' whose difficulty is measured directly relative to a model's current competence rather than using static proxy scores.
Paradigm Shift arxiv | Mar 17
KoopmanFlow uses a Koopman-inspired structural bias to decouple global steady-state motions from high-frequency local corrections in robotic control policies.
New Capability arxiv | Mar 17
Groups with bounded rationality and stochasticity can outperform perfectly rational agents because randomness encodes signals lost in deterministic behavior.
Breaks Assumption arxiv | Mar 17
Traditional Spiking Neural Network (SNN) sparsity is a performance 'illusion' on GPUs; temporal aggregation is required for actual 13x speedups.
Efficiency Breakthrough arxiv | Mar 17
ImagiNav enables robots to learn navigation from diverse 'in-the-wild' internet videos by decoupling visual planning from physical actuation.
Paradigm Shift arxiv | Mar 17
EVE rethinks neural architecture by replacing scalar units with local variational probabilistic neurons.
Paradigm Shift arxiv | Mar 17
GradMem replaces the massive KV-cache with a compact memory state updated via test-time gradient descent.
New Capability arxiv | Mar 17
A massive study of 19 LLMs reveals that subtle identity cues in names and dialects systematically bias automated text annotation.
Breaks Assumption arxiv | Mar 17
Redefines robotic visual state representations by explicitly encoding 'what-is-where' composition through a global-to-local reconstruction objective.
Paradigm Shift arxiv | Mar 17
Provides empirical evidence that LLMs hallucinate not from a lack of internal uncertainty, but because that uncertainty is 'functionally silent' during output generation.
Breaks Assumption arxiv | Mar 17
Reformulates traditional vision tasks like classification and object detection as a continuous transport process using Discriminative Flow Matching.
Paradigm Shift arxiv | Mar 17
Enables training of CNNs from scratch in true 4-bit precision on commodity CPUs with virtually no loss in accuracy.
Efficiency Breakthrough arxiv | Mar 17
Introduces a unified evaluation harness for Vision-Language-Action (VLA) models that standardizes disparate protocols and exposes hidden flaws in published SOTA models.
Open Release arxiv | Mar 17
Introduces the FLUX preprocessing pipeline, which reduces LLM training compute by 34% by maximizing high-quality token retention.
Efficiency Breakthrough arxiv | Mar 17
Reduces the RAM requirement for speech neuroprosthesis CTC decoding from 320 GB to 10 GB without sacrificing accuracy.
Efficiency Breakthrough arxiv | Mar 17
Proposes URDF-Anything+, an autoregressive framework that generates fully executable articulated 3D models from raw visual observations.
New Capability arxiv | Mar 17
Introduces the first system capable of imaging high-speed, non-rigid objects through strong atmospheric turbulence at 16,000 pixels per second.
New Capability arxiv | Mar 17
Enhances mathematical reasoning in LLMs by integrating Group Relative Policy Optimization (GRPO) with a specific reflection reward mechanism.
Paradigm Shift arxiv | Mar 17
Reveals that Graph-RAG performance is limited by reasoning failure rather than retrieval, and shows how to make an 8B model match a 70B baseline.
Efficiency Breakthrough arxiv | Mar 17
Amortizes iterative diffusion into a one-step trajectory policy for robotics using a novel 'Keyed Drift Field' objective.
Efficiency Breakthrough arxiv | Mar 17
Proposes a temporal mixed-precision framework for diffusion models that adaptively assigns bitwidths across different denoising timesteps.
Efficiency Breakthrough arxiv | Mar 17
Identifies a structural flaw in the standard Expected Calibration Error (ECE) when applied to soft labels and introduces SMECE to fix it.
Breaks Assumption arxiv | Mar 17
Accelerates LLM inference by up to 1.8x using a training-free sparse pattern predictor based on SVD truncation of FFN gate matrices.
Efficiency Breakthrough arxiv | Mar 17
Challenges the monotonic 'bigger is better' scaling paradigm by proving that institutional fitness peaks at an environment-dependent scale.
Scaling Insight arxiv | Mar 17
Introduces Centered Reward Distillation (CRD) to stabilize diffusion reinforcement learning by removing intractable normalizing constants.
Paradigm Shift arxiv | Mar 17
Demonstrates that gated predictive autoencoders can match or outperform JEPA-style architectures by learning to select predictable components.
Breaks Assumption arxiv | Mar 17
Unifies KV cache compression and sparse attention into a single 1-bit indexing structure, eliminating the need for external metadata or predictors.
Efficiency Breakthrough arxiv | Mar 17
Enables online, incremental 3D Gaussian Splatting for thousands of frames by replacing global reprocessing with a causal, streaming update framework.
New Capability arxiv | Mar 17
Detects diffusion-generated images 126x faster than reconstruction-based methods by using Gaussian noise disturbance to exploit the statistical 'ease' of fitting synthetic data.
Efficiency Breakthrough arxiv | Mar 17
Identifies that extended reasoning in Multimodal LLMs causes 'attention dispersion,' where models literally lose focus on visual inputs as the reasoning chain lengthens.
Breaks Assumption arxiv | Mar 17
Enables model adaptation on edge devices and non-differentiable (quantized) models using a purely backpropagation-free optimization framework.
Efficiency Breakthrough arxiv | Mar 17
Discovers that frozen video diffusion models already encode physical plausibility in their features, allowing for cost-effective inference-time physics filtering.
Breaks Assumption arxiv | Mar 17
Introduces a decentralized, multi-agent framework for scientific discovery that uses an 'ArtifactReactor' for plannerless coordination and full computational lineage.
New Capability arxiv | Mar 17
Proposes spectral clipping to stabilize LLM training by addressing 'spectral spikes' in stochastic gradient noise that adaptive optimizers like AdamW fail to handle.
Scaling Insight arxiv | Mar 17
Achieves real-time, low-latency talking avatar generation at 34ms per frame using a one-step streaming diffusion framework.
Efficiency Breakthrough arxiv | Mar 17
Introduces Matrix-to-Matrix RNNs (M$^2$RNN) with matrix-valued hidden states that outperform hybrid Transformers while using 3x smaller state sizes.
Scaling Insight arxiv | Mar 17
Proposes the 'Theory Compiler,' a system that automatically translates formal domain specifications into neural architectures with built-in physical or logical constraints.
Paradigm Shift arxiv | Mar 17
Introduces 'Visual Chronometer' to estimate physical frame rates directly from visual dynamics, addressing the 'chronometric hallucinations' common in generative video models.
New Capability arxiv | Mar 17
Segment Anything Reasoner (StAR) successfully introduces parallel test-time scaling to visual segmentation tasks, eliciting latent reasoning capabilities from base models.
New Capability arxiv | Mar 17
Argues that probability gradients are superior to standard log-probability gradients for RL training, proposing a new optimization method (DGPO) to solve divergence in soft clipping.
Breaks Assumption arxiv | Mar 17
Presents DataEvolve, a framework that enables AI to autonomously evolve and iteratively optimize pretraining data curation strategies.
Paradigm Shift arxiv | Mar 17
Introduces ZoomUI, a trainless method for GUI grounding that uses inference-time scaling to anchor natural language instructions to interface elements.
Efficiency Breakthrough arxiv | Mar 17
FLORE achieves 1000x error reduction in linear sketching while being 100x faster than previous learning-based solutions.
Efficiency Breakthrough arxiv | Mar 17
V-JEPA 2.1 unlocks dense, spatially structured features in video self-supervised learning, yielding massive gains in robotic manipulation and navigation.
New Capability arxiv | Mar 17
This paper provides a new identifiability theorem for causal representation learning to uncover physical system parameters from raw data without predefined libraries.
Paradigm Shift arxiv | Mar 17
The Infinite Problem Generator (IPG) uses executable code to synthesize and verify 100% accurate physics reasoning data, overcoming LLM hallucination in data scaling.
Scaling Insight arxiv | Mar 17
Simple regularization and data-hybrid training are shown to be sufficient to prevent catastrophic forgetting in MLLMs, challenging the need for complex anti-forgetting architectures.
Breaks Assumption arxiv | Mar 17
SleepGate introduces a biologically inspired 'sleep cycle' for the KV cache to resolve proactive interference in long-context LLMs.
Efficiency Breakthrough arxiv | Mar 17
One-Policy-Fits-All (OPFA) learns a single manipulation policy across 11 different embodiments, including grippers and dexterous hands, using geometry-aware action latents.
New Capability arxiv | Mar 17
Interp3R is the first method to estimate depth and camera poses at arbitrary time instants by interpolating pointmaps using asynchronous event data.
New Capability arxiv | Mar 17
Distilled VAE encoders are found to perform significantly better on higher, unseen resolutions than on their native training resolution.
Breaks Assumption arxiv | Mar 17
ASAP reduces LVLM computational FLOPs by ~80% with virtually no loss in performance using a training-free KV-Cache pruning recipe.
Efficiency Breakthrough arxiv | Mar 17
MorFiC achieves zero-shot locomotion transfer across quadrupeds of different sizes and masses with up to 5x speed gains over standard baselines.
New Capability arxiv | Mar 17
Top-b sampling introduces entropy-aware adaptive bandwidth for LLM decoding, effectively approximating a self-regulating control system for generation.
Paradigm Shift arxiv | Mar 17
SuperLocalMemory V3 establishes information-geometric foundations for agent memory, enabling high-accuracy retrieval without cloud-based LLM dependency.
Paradigm Shift arxiv | Mar 17
FlashHead is a drop-in replacement for the LM classification head that provides 1.75x inference speedup by treating vocabulary selection as a retrieval problem.
Efficiency Breakthrough arxiv | Mar 17
Introduces 'Delight' to policy gradients, weighting updates by the product of advantage and action surprisal to fix pathologies in RL training.
Paradigm Shift arxiv | Mar 17
Determines the optimal compute distribution for retrieval agents, showing that re-ranking depth is far more critical than query expansion strength.
Scaling Insight arxiv | Mar 17
Proposes the Spectrum Matching Hypothesis to explain why some VAE latents are 'undiffusable' and introduces techniques to align power spectral densities for superior image generation.
Paradigm Shift arxiv | Mar 17
Discovers interpretable 'atoms' of model behavior by decomposing training gradients, enabling unsupervised discovery and steering of complex behaviors like refusal or arithmetic.
New Capability arxiv | Mar 17