4 papers
KVarN: Variance-Normalized KV-Cache Quantization Mitigates Error Accumulation in Reasoning Tasks
Lorenz K. Muller, Philippe Bich, Chiara Boretti +3
Test-time scaling is a powerful approach to obtain better reasoning in large language models, but it becomes memory-bottlenecked during long-horizon decoding, as the KV-cache grows…
Linearly-Interpretable Concept Embedding Models for Text Analysis
Francesco De Santis, Philippe Bich, Gabriele Ciravegna +3
Despite their success, Large-Language Models (LLMs) still face criticism due to their lack of interpretability. Traditional post-hoc interpretation methods, based on attention and…
Towards Better Generalization and Interpretability in Unsupervised Concept-Based Models
Francesco De Santis, Philippe Bich, Gabriele Ciravegna +3
To increase the trustworthiness of deep neural networks, it is critical to improve the understanding of how they make decisions. This paper introduces a novel unsupervised concept-…
V-CEM: Bridging Performance and Intervenability in Concept-based Models
Francesco De Santis, Gabriele Ciravegna, Philippe Bich +2
Concept-based eXplainable AI (C-XAI) is a rapidly growing research field that enhances AI model interpretability by leveraging intermediate, human-understandable concepts. This app…