5 papers
PRISM: Transfer and Limits of Activation Interventions in Tool-Calling Language Models
Giovana Kerche Bonás, Thales Sales Almeida, Marcelo da Silva Reis +2
Recent literature shows that unsafe tool calls can often accompany textual refusals in language models. To investigate this hidden vulnerability, we introduce PRISM, a framework fo…
Detecting Hallucination in LLMs: Tracing the Topological Signatures of Impaired Context Sharing
Amir Jalilifard, Anderson Rocha, Eric Wong +1
In this work, we examine the topology of information flow patterns within attention graphs to effectively distinguish hallucinated from non-hallucinated responses. We analyze the F…
Reasoning Distillation and Structural Alignment for Improved Code Generation
Amir Jalilifard, Anderson de Rezende Rocha, Marcos Medeiros Raimundo
Effective code generation with language models hinges on two critical factors: accurately understanding the intent of the prompt and generating code that applies algorithmic reason…
Multi-criteria Rank-based Aggregation for Explainable AI
Sujoy Chatterjee, Everton Romanzini Colombo, Marcos Medeiros Raimundo
Explainability is crucial for improving the transparency of black-box machine learning models. With the advancement of explanation methods such as LIME and SHAP, various XAI perfor…
Pareto sensitivity, most-changing sub-fronts, and knee solutions
Tommaso Giovannelli, Marcos Medeiros Raimundo, Luis Nunes Vicente
When dealing with a multi-objective optimization problem, identifying Pareto optimal solutions that capture relevant trade-offs may not be straightforward. Popular choices include…