PARADIGM_SHIFT PARADIGM_SHIFT
329 papers · Page 4 of 4
Switches the training objective from hard Next-Token Prediction to predicting 'concepts' (sets of semantically related tokens).
AI & ML arxiv | Apr 1
Proves that LLM agent capability (pass@1) and reliability (consistency) diverge systematically, with frontier models often having the highest 'meltdown' rates.
AI & ML arxiv | Apr 1
Learns stable, interpretable Koopman generators for nonlinear PDEs from trajectory data alone without any physics supervision.
AI & ML arxiv | Apr 1
Shows that VLMs can overcome deep-seated perceptual biases and optical illusions by using image manipulation tools rather than more training data.
AI & ML arxiv | Apr 1
A novel neural primitive based on metriplectic dynamics that outperforms Transformers in data efficiency and generalization.
AI & ML arxiv | Apr 1
A unified agentic framework that closes the 'AI-for-AI' research loop by discovering novel architectures, data pipelines, and algorithms.
AI & ML arxiv | Apr 1
Decouples high-level intent planning from low-level motor control in Vision-Language-Action (VLA) models to prevent the degradation of pre-trained VLM representations.
AI & ML arxiv | Apr 1
Demonstrates that independent aggregation (Hybrid Confirmation Tree) consistently outperforms the standard 'AI-as-advisor' paradigm across diverse high-stakes domains.
AI & ML arxiv | Apr 1
Shows that deep learning models for medical imaging (MRI) can be trained using synthetic quaternion Julia fractals instead of sensitive human clinical data.
AI & ML arxiv | Apr 1
Provides a formal framework for optimizing models whose decisions actively change the distribution of the data they encounter.
AI & ML arxiv | Apr 1
Introduces a rigorous algorithm to determine if two different neural networks share the same underlying 'algorithmic interpretation' without needing to manually define the circuits.
AI & ML arxiv | Apr 1
Replaces heuristic ReAct-style agent loops with a mathematical framework based on control theory to prevent LLM agents from over-deliberating or using excessive tools.
AI & ML arxiv | Apr 1
First foundation model to unify text, image, audio, and video using native masked diffusion instead of autoregressive serialization.
AI & ML arxiv | Apr 2
LLM-guided program evolution has discovered a new data-shuffling rule for SGD that provably and empirically outperforms standard Random Reshuffling.
AI & ML arxiv | Apr 2
A comprehensive analysis of AI safety vulnerabilities including automated circuit discovery, latent adversarial training, and power-law scaling of jailbreak success.
AI & ML arxiv | Apr 2
Identifies a fundamental quality-exploration dilemma in Diffusion Language Models where remasking improves single-sample quality but kills reasoning diversity.
AI & ML arxiv | Apr 2
Introduces training-free and model-free trajectory planning by computing diffusion score functions directly from data libraries via kernel-weighted estimation.
AI & ML arxiv | Apr 2
Proposes a decision-centric architecture that separates signal estimation from control policy to make LLM system decisions explicit and inspectable.
AI & ML arxiv | Apr 2
Truth Anchoring (TAC) provides a post-hoc calibration method to align LLM uncertainty metrics with actual factual correctness.
AI & ML arxiv | Apr 2
Identifies 'diversity collapse' in the popular GRPO reinforcement learning method and introduces MUPO to maintain broad reasoning paths.
AI & ML arxiv | Apr 2
Replaces manual rubric-tuning for synthetic data with an automated gradient-guided optimization framework based on influence estimation.
AI & ML arxiv | Apr 2
Introduces HiLL, a framework that jointly trains a 'hinter' and 'reasoner' to prevent advantage collapse in reinforcement learning for hard tasks.
AI & ML arxiv | Apr 2
LangMARL introduces agent-level credit assignment and policy gradient evolution directly in the natural language space for multi-agent coordination.
AI & ML arxiv | Apr 2
Stochastic Attention achieves a global receptive field in O(log n) layers by using randomized routing inspired by the fruit fly connectome.
AI & ML arxiv | Apr 2
Routing-Free MoE replaces centralized routing with individual expert-level activation, eliminating the need for Softmax and Top-K load balancing.
AI & ML arxiv | Apr 2
Policy Improvement Reinforcement Learning (PIRL) shifts the training objective from reward maximization to explicit maximization of policy progress across iterations.
AI & ML arxiv | Apr 2
Proposes dense point trajectories as universal 'visual tokens' for behavior that generalize across different species and non-rigid objects.
AI & ML arxiv | Apr 2
Achieves 'zero forgetting' in continual learning by stacking frozen domain-specific MoE-LoRA adapters with a meta-router.
AI & ML arxiv | Apr 2
Replaces standard relative Softmax attention with 'Multiscreening' to allow absolute query-key relevance, yielding 3.2x faster inference at 100K context.
AI & ML arxiv | Apr 2