11 papers
KV Cache Compression Through the Lens of Transform Coding
Hannah Laus, Claudio Mayrink Verdun, Hao Wang +2
The key-value (KV) cache stores information from past tokens and is a major memory bottleneck in long-context inference. Existing quantization methods address this bottleneck by re…
Inference-Time Machine Unlearning via Gated Activation Redirection
VinÃcius Conte Turani, Otávio Parraga, João Vitor Boer Abitante +7
The paper proposes GUARD-IT, a gradient‑free method that modifies activations at inference time with input‑dependent rotations to erase specific data from large language models whi…
Gaussian DP for Reporting Differential Privacy Guarantees in Machine Learning
Juan Felipe Gomez, Bogdan Kulynych, Georgios Kaissis +4
Current practices for reporting differential privacy (DP) guarantees for machine learning (ML) algorithms such as DP-SGD provide an incomplete and potentially misleading picture. F…
Optimal conversion from Rényi Differential Privacy to -Differential Privacy
Anneliese Riess, Juan Felipe Gomez, Flavio du Pin Calmon +2
We prove the conjecture stated in Appendix F.3 of \citet{zhu2022optimalaccountingdifferentialprivacy}: among all conversion rules that map a Rényi Differential Privacy (RDP) profi…
Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management
Carol Xuan Long, David Simchi-Levi, Feng Zhu +3
This paper studies autonomous generative AI agents in multi-echelon supply chains using the MIT Beer Game. We identify four inference-time levers that shape performance: model sele…
Robust AI Evaluation through Maximal Lotteries
Hadi Khalaf, Serena L. Wang, Daniel Halpern +3
The standard way to evaluate language models on subjective tasks is through pairwise comparisons: an annotator chooses the "better" of two responses to a prompt. Leaderboards aggre…