The Kannan--Lov\'asz--Simonovits (KLS) conjecture asserts that isotropic log-concave probability measures have Poincar\'e constants bounded by a universal constant, independently of dimension. We give a deterministic variational proof with an explicit bound on the Poincar\'e constant $C_P(\mu)\le25$, where $\mu$ is any...
K. Balasubramanian, S. Kasiviswanathan· 0 citations
KV-cache entries are stored before their future queries are known, but each decoding query needs precision in different places. We study this mismatch using separate budgets for retained bits and bits fetched per query. ReadKV stores each key and value in a progressive code whose prefixes support different reconstructi...
Siddharth Bhandari, Lucas Gretta, K. Balasubramanian et al.· 0 citations
Counterfactual generation seeks to sample outcomes under a hypothetical intervention or decision using observational data collected under the factual assignment mechanism. We develop a flow-matching approach that combines a sample-split, doubly robust training objective with a learned coupling between observed source o...
Yun-Rui Guan, K. Balasubramanian, S. Kasiviswanathan· 0 citations
Group Adaptive Clipping Policy Optimization (GAPO), a plug-in modification to GRPO methods that adapts the clipping boundary to the rollout advantage, is proposed, motivated by a reverse-KL trust-region perspective, which suggests that rollouts with larger learning signal should receive proportionally greater update he...
Sheng Jia, Xiao Wang, S. Kasiviswanathan et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
