archive
Every paper Pith has read. Search by title, abstract, or pith.
554 papers in cs.NE · page 6
-
Binary spiking LM matches performance at 5 percent compute
BiSpikCLM: A Spiking Language Model integrating Softmax-Free Spiking Attention and Spike-Aware Alignment Distillation
-
Wolf-inspired agents optimize LLM prompts and hyperparameters
Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language Models
-
Trainable parameters enable adaptive firing in spiking neurons
Adaptive Spiking Neurons for Vision and Language Modeling
-
Meta-learning speeds evolutionary adaptation to drifting data streams
GeM-EA: A Generative and Meta-learning Enhanced Evolutionary Algorithm for Streaming Data-Driven Optimization
-
Auxiliary unknown class unifies calibration and classification training
Socrates Loss: Unifying Confidence Calibration and Classification by Leveraging the Unknown
-
Regular firing pattern shields SNN weights from forgetting
Gradient-Free Continual Learning in Spiking Neural Networks via Inter-Spike Interval Regularization
-
Spiking network triggers LLM action after idle
EMBER: Autonomous Cognitive Behaviour from Learned Spiking Neural Network Dynamics in a Hybrid LLM Architecture
-
Hybrid learning in spiking nets monitors nuclear plants sequentially
Neuromorphic Continual Learning for Sequential Deployment of Nuclear Plant Monitoring Systems
-
AI life detectors fooled by simulated non-life
Can AI Detect Life? Lessons from Artificial Life
-
Inverted Galois algebra yields STDP-like selection in HDC
Beyond LLMs, Sparse Distributed Memory, and Neuromorphics <A Hyper-Dimensional SRAM-CAM "VaCoAl" for Ultra-High Speed, Ultra-Low Power, and Low Cost>
-
Inverted Galois algebra yields STDP-like selection in HDC
Beyond LLMs, Sparse Distributed Memory, and Neuromorphics <A Hyper-Dimensional SRAM-CAM "VaCoAl" for Ultra-High Speed, Ultra-Low Power, and Low Cost>
-
SpikeMLLM keeps MLLM accuracy within 1% at three-quarters timestep count
SpikeMLLM: Spike-based Multimodal Large Language Models via Modality-Specific Temporal Scales and Temporal Compression
-
Hormone block adds continuous emotion to T5 transformers
A Hormone-inspired Emotion Layer for Transformer language models (HELT)
-
Winner-take-all attention powers spiking language transformers
Winner-Take-All Spiking Transformer for Language Modeling
-
Neural cellular automata evolve diverse persistent lifelike behaviors
Evolving Many Worlds: Towards Open-Ended Discovery in Petri Dish NCA via Population-Based Training
-
Local LLMs build knowledge graphs zero-shot at 0.70 F1
Frugal Knowledge Graph Construction with Local LLMs: A Zero-Shot Pipeline, Self-Consistency and Wisdom of Artificial Crowds
-
K-way energy probes reduce to softmax in PC networks
K-Way Energy Probes for Metacognition Reduce to Softmax in Discriminative Predictive Coding Networks
-
MOEAs display distinct behaviors on stochastic dynamic knapsacks
On the Use of Bi-Objective Evolutionary Algorithms for the Stochastic MKP under Dynamic Constraints
-
ResNet segments retinal cysts above 70% Dice across vendors
Retinal Cyst Detection from Optical Coherence Tomography Images
-
Self-certifying cache proves its own error bound at runtime
LAWS: Learning from Actual Workloads Symbolically -- A Self-Certifying Parametrized Cache Architecture for Neural Inference, Robotics, and Edge Deployment
-
Transformer sets its own head count via online geometric checks
INCRT: An Incremental Transformer That Determines Its Own Architecture
-
Macroscopic views simplify neural cellular automata manifolds
Visualising the Attractor Landscape of Neural Cellular Automata
-
TurboEvolve raises LLM program quality at lower evaluation budgets
TurboEvolve: Towards Fast and Robust LLM-Driven Program Evolution
-
AI architecture changes match biological mutation statistics
Universal statistical signatures of evolution in artificial intelligence architectures
-
Optimized hub placement improves sparse training
Heterogeneous Connectivity in Sparse Networks: Fan-in Profiles, Gradient Hierarchy, and Topological Equilibria
-
Spike-driven LLM cuts energy 7x while raising accuracy 4.2%
Spike-driven Large Language Model
-
Closed-form bound derived for Hessian max eigenvalue in smooth NNs
Wolkowicz-Styan Upper Bound on the Hessian Eigenspectrum for Cross-Entropy Loss in Nonlinear Smooth Neural Networks
-
Microcontroller runs full SNN simulation at 20 mW
Full Feature Spiking Neural Network Simulation on Micro-Controllers for Neuromorphic Applications at the Edge
-
KV caches compress 914000x better than per-vector methods
Sequential KV Cache Compression via Probabilistic Language Tries: Beyond the Per-Vector Shannon Limit
-
Genetic algorithm evolves token vectors to optimize diffusion prompts
Evolutionary Token-Level Prompt Optimization for Diffusion Models
-
No one material leads physical neural computing across all measures
Beyond Silicon: Materials, Mechanisms, and Methods for Physical Neural Computing
-
Online framework sustains accuracy on drifting time series without labels
Drift-Aware Online Dynamic Learning for Nonstationary Multivariate Time Series: Application to Sintering Quality Prediction
-
0.5V encoder maps voltages to spikes within 5.6 percent linearity
A 0.5-V Linear Neuromorphic Voltage-to-Spike Encoder Using a Bulk-Driven Transconductor
-
King Wen sequence degrades neural network training
Statistical Properties of the King Wen Sequence: An Anti-Habituation Structure That Does Not Improve Neural Network Training
-
VENs speed social decisions via a fast pathway
The Fast Lane Hypothesis: Von Economo Neurons Implement a Biological Speed-Accuracy Tradeoff
-
Simulations show agents building social groups via conformity and creativity
Social Reality Construction via Active Inference: Modeling the Dialectic of Conformity and Creativity
-
Multi-dimensional grouping slashes energy in spiking transformers
Ge$^\text{2}$mS-T: Multi-Dimensional Grouping for Ultra-High Energy Efficiency in Spiking Transformer
-
Hierarchical attention subsumes standard attention at 1.31x cost
Hierarchical Kernel Transformer: Multi-Scale Attention with an Information-Theoretic Approximation Analysis
-
Neuromorphic chips hit new memory wall from on-chip storage
Memory Wall is not gone: A Critical Outlook on Memory Architecture in Digital Neuromorphic Computing
-
Frozen random backbones with LoRA match full training
A Little Rank Goes a Long Way: Random Scaffolds with LoRA Adapters Are All You Need
-
SNIP alignment fails to improve during symbolic regression optimization
Multi-Modal Learning meets Genetic Programming: Analyzing Alignment in Latent Space Optimization
-
This paper develops a tri-objective genetic algorithm for planning overnight truck routes…
Robust Multi-Objective Optimization for Bicycle Rebalancing in Shared Mobility Systems
-
Echo Networks encode neural nets as one matrix for matrix-based evolution
Introducing Echo Networks for Computational Neuroevolution
-
Bijective transformations shift multi-objective algorithm rankings
Exploration of Pareto-preserving Search Space Transformations in Multi-objective Test Functions
-
Activation before noise injection raises neural net accuracy
Internal noise in deep neural networks: interplay of depth, neuron number, and noise injection step
-
EA selects Condorcet winner constantly often when edge is O(1/n)
Analysis of Search Heuristics in the Multi-Armed Bandit Setting
-
Phase synchronization boosts vision-transformer efficiency
Kuramoto Oscillatory Phase Encoding: Neuro-inspired Synchronization for Improved Learning Efficiency
-
Metacognitive loop lets LLMs evolve stronger VRP heuristics
PyVRP$^+$: LLM-Driven Metacognitive Heuristic Evolution for Hybrid Genetic Search in Vehicle Routing Problems
-
TrilinearCIM runs Transformer attention in NVM without reprogramming
Trilinear Compute-in-Memory Architecture for Energy-Efficient Transformer Acceleration
-
Evolutionary search yields Pareto sets of forecasting networks
Auto-Configured Networks for Multi-Scale Multi-Output Time-Series Forecasting