FERA improves LLM multi-step reasoning in federated settings via iterative uncertainty-aware aggregation of client traces without data sharing or training.
Context: You are given several reasoning responses to the same question, each from a different client
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
FERA: Uncertainty-Aware Federated Reasoning for Large Language Models
FERA improves LLM multi-step reasoning in federated settings via iterative uncertainty-aware aggregation of client traces without data sharing or training.