Image de couverture de livre

Hidden Breakthroughs in Language Model Training

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Kangaslahti, Sara, Rosenfeld, Elan, Saphra, Naomi
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Fast Forwarding Low-Rank Training
par: Rahamim, Adir, et autres
Publié: (2024)

One-Shot Strategic Classification Under Unknown Costs
par: Rosenfeld, Elan, et autres
Publié: (2023)

Latent State Models of Training Dynamics
par: Hu, Michael Y., et autres
Publié: (2023)

Continuous Language Model Interpolation for Dynamic and Controllable Text Generation
par: Kangaslahti, Sara, et autres
Publié: (2024)

Mechanistic?
par: Saphra, Naomi, et autres
Publié: (2024)

Sometimes I am a Tree: Data Drives Unstable Hierarchical Generalization
par: Qin, Tian, et autres
Publié: (2024)

PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs
par: van der Wal, Oskar, et autres
Publié: (2025)

Deep sequence models tend to memorize geometrically; it is unclear why
par: Noroozizadeh, Shahriar, et autres
Publié: (2025)

TRAM: Bridging Trust Regions and Sharpness Aware Minimization
par: Sherborne, Tom, et autres
Publié: (2023)

Identifying Representations for Intervention Extrapolation
par: Saengkyongam, Sorawit, et autres
Publié: (2023)

Transcendence: Generative Models Can Outperform The Experts That Train Them
par: Zhang, Edwin, et autres
Publié: (2024)

World Properties without World Models: Recovering Spatial and Temporal Structure from Co-occurrence Statistics in Static Word Embeddings
par: Barenholtz, Elan
Publié: (2026)

Inverse Depth Scaling From Most Layers Being Similar
par: Liu, Yizhou, et autres
Publié: (2026)

Boomerang Distillation Enables Zero-Shot Model Size Interpolation
par: Kangaslahti, Sara, et autres
Publié: (2025)

Random Scaling of Emergent Capabilities
par: Zhao, Rosie, et autres
Publié: (2025)

Do Activation Verbalization Methods Convey Privileged Information?
par: Li, Millicent, et autres
Publié: (2025)

Can Interpretation Predict Behavior on Unseen Data?
par: Li, Victoria R., et autres
Publié: (2025)

Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention
par: Huang, Jing, et autres
Publié: (2026)

Analyzing Political Text at Scale with Online Tensor LDA
par: Kangaslahti, Sara, et autres
Publié: (2025)

Learning under noisy supervision is governed by a feedback-truth gap
par: Schonfeld, Elan, et autres
Publié: (2026)

Attribute Diversity Determines the Systematicity Gap in VQA
par: Berlot-Attwell, Ian, et autres
Publié: (2023)

Using Shapley interactions to understand how models use structure
par: Singhvi, Divyansh, et autres
Publié: (2024)

K-Edit: Language Model Editing with Contextual Knowledge Awareness
par: Markowitz, Elan, et autres
Publié: (2025)

FastSurvival: Hidden Computational Blessings in Training Cox Proportional Hazards Models
par: Liu, Jiachang, et autres
Publié: (2024)

Strategic Classification with Non-Linear Classifiers
par: Trachtenberg, Benyamin, et autres
Publié: (2025)

Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration
par: Yu, Zhongzhi, et autres
Publié: (2024)

Instructed to Bias: Instruction-Tuned Language Models Exhibit Emergent Cognitive Bias
par: Itzhak, Itay, et autres
Publié: (2023)

Exploring the Hidden Reasoning Process of Large Language Models by Misleading Them
par: Chen, Guanyu, et autres
Publié: (2025)

Embedding Hidden Adversarial Capabilities in Pre-Trained Diffusion Models
par: Beerens, Lucas, et autres
Publié: (2025)

Multiway Multislice PHATE: Visualizing Hidden Dynamics of RNNs through Training
par: Xie, Jiancheng, et autres
Publié: (2024)

Transfer Learning in Robotics: An Upcoming Breakthrough? A Review of Promises and Challenges
par: Jaquier, Noémie, et autres
Publié: (2023)

Single-Rollout Hidden-State Dynamics for Training-Free RLVR Data Selection
par: Wu, Jianghao, et autres
Publié: (2026)

Pre-trained Large Language Models Learn Hidden Markov Models In-context
par: Dai, Yijia, et autres
Publié: (2025)

Latent Spherical Flow Policy for Reinforcement Learning with Combinatorial Actions
par: Kong, Lingkai, et autres
Publié: (2026)

Prototype Training with Dual Pseudo-Inverse and Optimized Hidden Activations
par: Tucci, Mauro
Publié: (2025)

Unveiling Hidden Collaboration within Mixture-of-Experts in Large Language Models
par: Tang, Yuanbo, et autres
Publié: (2025)

Discovering and Reasoning of Causality in the Hidden World with Large Language Models
par: Liu, Chenxi, et autres
Publié: (2024)

Layer by Layer: Uncovering Hidden Representations in Language Models
par: Skean, Oscar, et autres
Publié: (2025)

A Market for Accuracy: Classification under Competition
par: Einav, Ohad, et autres
Publié: (2025)

Welfare as a Guiding Principle for Machine Learning -- From Compass, to Lens, to Roadmap
par: Rosenfeld, Nir, et autres
Publié: (2025)