On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

169,051 papers · 148 categories

Trend · papers per month

count % of corpus

4 results for “Gumbel-Top-k”

Stochastic Beams and Where to Find Them: The Gumbel-Top-k Trick for Sampling Sequences Without Replacementcs.LG

A new method samples sequences without replacement using Gumbel-Top-k trick.

problem Sampling sequences without replacement from large domains efficiently.

method Gumbel-Top-k trick applied to Stochastic Beam Search.

result Efficient sampling with linear model evaluations in

k

sampling sequences Gumbel-Top-k Stochastic Beam Search diversity

2019-03-14abs ↗pdf ↗

Low-variance estimation in the Plackett-Luce model via quasi-Monte Carlo samplingstat.ML

New method reduces variance in estimating PL model expectations.

problem High variance in Monte Carlo estimates of PL model expectations.

method Combining Gumbel top-k trick with quasi-Monte Carlo sampling.

result More sample-efficient estimators of PL model expectations.

Plackett-Luce model ranking metrics Monte Carlo sampling quasi-Monte Carlo variance reduction

2022-05-12abs ↗pdf ↗

Rank-Conditioned Sample Reuse for the Plackett--Luce Best-of-

K

Objectivecs.LG

New method uses rank-conditioned Horvitz-Thompson estimation for unbiased sample reuse in Plackett-Luce best-of-K objective.

problem Estimating the expected maximum reward in Plackett-Luce draws without replacement.

method Rank-conditioned Horvitz-Thompson estimation with joint-score REINFORCE for unbiased sample reuse.

result Unbiased estimation of the Plackett-Luce best-of-K objective with finite second moment guarantees.

Plackett-Luce sample reuse best-of-K Horvitz-Thompson REINFORCE

2026-07-13abs ↗pdf ↗

Arithmetic Sampling: Parallel Diverse Decoding for Large Language Modelscs.CL

We introduce a new sampling method for large language models that balances diversity and parallelism.

problem Balancing diversity and parallelism in decoding for large language models.

method Arithmetic sampling framework compatible with various sampling variations.

result Improves estimation of expected BLEU score reward and reduces the gap with beam search.

large language models beam search diversity parallelism arithmetic sampling

2022-10-18abs ↗pdf ↗