Cover Image

Compositional Generalization in Autoregressive Models via Logit Composition

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Kumar, Aakash, Bucarelli, Maria Sofia, Natale, Emanuele
Format:	Preprint
Published:	2026
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Quantization vs Pruning: Insights from the Strong Lottery Ticket Hypothesis
by: Kumar, Aakash, et al.
Published: (2025)

Beyond Fixed Points: Superpolynomial Capacity of Asymmetric Hopfield Networks
by: Kumar, Aakash, et al.
Published: (2026)

ATM: Improving Model Merging by Alternating Tuning and Merging
by: Zhou, Luca, et al.
Published: (2024)

Learning with Noisy Labels through Learnable Weighting and Centroid Similarity
by: Wani, Farooq Ahmad, et al.
Published: (2023)

Logit-KL Flow Matching: Non-Autoregressive Text Generation via Sampling-Hybrid Inference
by: Sevriugov, Egor, et al.
Published: (2024)

PISA: Prioritized Invariant Subgraph Aggregation
by: Ghasemi, Ali, et al.
Published: (2025)

Subtract the Corruption: Training-Data-Free Corrective Machine Unlearning using Task Arithmetic
by: Mozafari, Mostafa, et al.
Published: (2025)

Task Singular Vectors: Reducing Task Interference in Model Merging
by: Gargiulo, Antonio Andrea, et al.
Published: (2024)

On Task Vectors and Gradients
by: Zhou, Luca, et al.
Published: (2025)

The Majority Vote Paradigm Shift: When Popular Meets Optimal
by: Purificato, Antonio, et al.
Published: (2025)

A topological description of loss surfaces based on Betti Numbers
by: Bucarelli, Maria Sofia, et al.
Published: (2024)

Compositional Capabilities of Autoregressive Transformers: A Study on Synthetic, Interpretable Tasks
by: Ramesh, Rahul, et al.
Published: (2023)

Logit Distance Bounds Representational Similarity
by: Nielsen, Beatrix M. G., et al.
Published: (2026)

Energy Guided smoothness to improve Robustness in Graph Classification
by: Wani, Farooq Ahmad, et al.
Published: (2024)

Link Prediction with Physics-Inspired Graph Neural Networks
by: Di Francesco, Andrea Giuseppe, et al.
Published: (2024)

Test-Time Compositional Generalization in Diffusion Models via Concept Discovery
by: Wang, Zekun, et al.
Published: (2026)

Early-Exit Graph Neural Networks
by: Di Francesco, Andrea Giuseppe, et al.
Published: (2025)

Enhancing Compositional Generalization via Compositional Feature Alignment
by: Wang, Haoxiang, et al.
Published: (2024)

Fast Graph Generation via Autoregressive Noisy Filtration Modeling
by: Krimmel, Markus, et al.
Published: (2025)

Polynomially Over-Parameterized Convolutional Neural Networks Contain Structured Strong Winning Lottery Tickets
by: da Cunha, Arthur, et al.
Published: (2023)

Trading-off Accuracy and Communication Cost in Federated Learning
by: Villani, Mattia Jacopo, et al.
Published: (2025)

MASS: MoErging through Adaptive Subspace Selection
by: Crisostomi, Donato, et al.
Published: (2025)

Contextual Multinomial Logit Bandits with General Value Functions
by: Zhang, Mengxiao, et al.
Published: (2024)

Consistency Regularization for Domain Generalization with Logit Attribution Matching
by: Gao, Han, et al.
Published: (2023)

$\nabla τ$: Gradient-based and Task-Agnostic machine Unlearning
by: Trippa, Daniel, et al.
Published: (2024)

A General Theory for Compositional Generalization
by: Fu, Jingwen, et al.
Published: (2024)

Online Continual Learning via Logit Adjusted Softmax
by: Huang, Zhehao, et al.
Published: (2023)

Detection Limits and Statistical Separability of Tree Ring Watermarks in Rectified Flow-based Text-to-Image Generation Models
by: Umrajkar, Ved, et al.
Published: (2025)

Discrete Compositional Generation via General Soft Operators and Robust Reinforcement Learning
by: Jiralerspong, Marco, et al.
Published: (2025)

Logits-Based Finetuning
by: Li, Jingyao, et al.
Published: (2025)

Structured vs. Unstructured Pruning: An Exponential Gap
by: Ferre', Davide, et al.
Published: (2026)

How Compositional Generalization and Creativity Improve as Diffusion Models are Trained
by: Favero, Alessandro, et al.
Published: (2025)

Scalable Evaluation and Neural Models for Compositional Generalization
by: Camposampiero, Giacomo, et al.
Published: (2025)

Compositional Generalization via Forced Rendering of Disentangled Latents
by: Liang, Qiyao, et al.
Published: (2025)

Stabilizing Policy Optimization via Logits Convexity
by: Chen, Hongzhan, et al.
Published: (2026)

Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
by: Li, Jin, et al.
Published: (2025)

From Logits to Hierarchies: Hierarchical Clustering made Simple
by: Palumbo, Emanuele, et al.
Published: (2024)

Active Exploration via Autoregressive Generation of Missing Data
by: Cai, Tiffany Tianhui, et al.
Published: (2024)

Individualised Treatment Effects Estimation with Composite Treatments and Composite Outcomes
by: Chauhan, Vinod Kumar, et al.
Published: (2025)

Logit Distillation on Manifolds: Mapping by Learning
by: Yang, Yiru, et al.
Published: (2026)