FlashEvaluator: Expanding Search Space with Parallel Evaluation

• URL: http://arxiv.org/abs/2603.02565v1
• Date: Tue, 03 Mar 2026 03:35:34 GMT
• Title: FlashEvaluator: Expanding Search Space with Parallel Evaluation
• Authors: Chao Feng, Yuanhao Pu, Chenghao Zhang, Shanqi Liu, Shuchang Liu, Xiang Li, Yongqi Liu, Lantao Hu, Kaiqiao Zhan, Han Li, Kun Gai,
• Abstract summary: We propose FlashEvaluator, which enables cross-sequence token information sharing and processes all sequences in a single forward pass.<n>This yields sublinear computational complexity that improves the system's efficiency and supports direct inter-sequence comparisons.<n>FlashEvaluator has been deployed in online recommender system of Kuaishou, delivering substantial and sustained revenue gains in practice.
• Score: 27.839033452409428
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The Generator-Evaluator (G-E) framework, i.e., evaluating K sequences from a generator and selecting the top-ranked one according to evaluator scores, is a foundational paradigm in tasks such as Recommender Systems (RecSys) and Natural Language Processing (NLP). Traditional evaluators process sequences independently, suffering from two major limitations: (1) lack of explicit cross-sequence comparison, leading to suboptimal accuracy; (2) poor parallelization with linear complexity of O(K), resulting in inefficient resource utilization and negative impact on both throughput and latency. To address these challenges, we propose FlashEvaluator, which enables cross-sequence token information sharing and processes all sequences in a single forward pass. This yields sublinear computational complexity that improves the system's efficiency and supports direct inter-sequence comparisons that improve selection accuracy. The paper also provides theoretical proofs and extensive experiments on recommendation and NLP tasks, demonstrating clear advantages over conventional methods. Notably, FlashEvaluator has been deployed in online recommender system of Kuaishou, delivering substantial and sustained revenue gains in practice.

Related papers

• $V_1$: Unifying Generation and Self-Verification for Parallel Reasoners [69.66089681814013]
  $V_$ is a framework that unifies generation and verification through efficient pairwise ranking.<n>$V_$-Infer improves Pass@1 by up to $10%$ over pointwise verification.<n>$V_$-PairRL achieves $7$--$9%$ test-time scaling gains over standard RL and pointwise joint training.
  arXiv  Detail & Related papers  (2026-03-04T17:22:16Z)
• Spend Search Where It Pays: Value-Guided Structured Sampling and Optimization for Generative Recommendation [16.991391135071513]
  We propose V-STAR, a Value-guided Sampling and Tree-structured Advantage Reinforcement framework.<n>V-STAR forms a self-evolving loop via two synergistic components. First, a Value-Guided Efficient Decoding (VED) is developed to identify decisive nodes and selectively deepen high-potential prefixes.<n>Second, we propose Sibling-GRPO, which exploits the induced tree topology to compute sibling-relative advantages and concentrates learning signals on decisive branching decisions.
  arXiv  Detail & Related papers  (2026-02-11T09:57:36Z)
• ArenaRL: Scaling RL for Open-Ended Agents via Tournament-based Relative Ranking [84.07076200941474]
  ArenaRL is a reinforcement learning paradigm that shifts from pointwise scalar scoring to intra-group relative ranking.<n>We construct an intra-group adversarial arena and devise a tournament-based ranking scheme to obtain stable advantage signals.<n>Experiments show that ArenaRL substantially outperforms standard RL baselines.
  arXiv  Detail & Related papers  (2026-01-10T08:43:07Z)
• DynScaling: Efficient Verifier-free Inference Scaling via Dynamic and Integrated Sampling [20.605487145370752]
  Inference-time scaling has proven effective in boosting large language model (LLM) performance through increased test-time computation.<n>Yet, its practical application is often hindered by reliance on external verifiers or a lack of optimization for realistic computational constraints.<n>We propose DynScaling, which addresses these limitations through two primary innovations: an integrated parallel-sequential sampling strategy and a bandit-based dynamic budget allocation framework.
  arXiv  Detail & Related papers  (2025-06-19T05:40:54Z)
• SPARE: Single-Pass Annotation with Reference-Guided Evaluation for Automatic Process Supervision and Reward Modelling [58.05959902776133]
  We introduce Single-Pass.<n>with Reference-Guided Evaluation (SPARE), a novel structured framework that enables efficient per-step annotation.<n>We demonstrate SPARE's effectiveness across four diverse datasets spanning mathematical reasoning (GSM8K, MATH), multi-hop question answering (MuSiQue-Ans), and spatial reasoning (SpaRP)<n>On ProcessBench, SPARE demonstrates data-efficient out-of-distribution generalization, using only $sim$16% of training samples compared to human-labeled and other synthetically trained baselines.
  arXiv  Detail & Related papers  (2025-06-18T14:37:59Z)
• On the Role of Feedback in Test-Time Scaling of Agentic AI Workflows [71.92083784393418]
  Agentic AI (systems that autonomously plan and act) are becoming widespread, yet their task success rate on complex tasks remains low.<n>Inference-time alignment relies on three components: sampling, evaluation, and feedback.<n>We introduce Iterative Agent Decoding (IAD), a procedure that repeatedly inserts feedback extracted from different forms of critiques.
  arXiv  Detail & Related papers  (2025-04-02T17:40:47Z)
• A Systematic Examination of Preference Learning through the Lens of Instruction-Following [83.71180850955679]
  We use a novel synthetic data generation pipeline to generate 48,000 instruction unique-following prompts.<n>With our synthetic prompts, we use two preference dataset curation methods - rejection sampling (RS) and Monte Carlo Tree Search (MCTS)<n>Experiments reveal that shared prefixes in preference pairs, as generated by MCTS, provide marginal but consistent improvements.<n>High-contrast preference pairs generally outperform low-contrast pairs; however, combining both often yields the best performance.
  arXiv  Detail & Related papers  (2024-12-18T15:38:39Z)
• Vector Optimization with Gaussian Process Bandits [7.049738935364297]
  Learning problems in which multiple objectives must be considered simultaneously often arise in various fields, including engineering, drug design, and environmental management.<n>Traditional methods for dealing with multiple black-box objective functions have limitations in incorporating objective preferences and exploring the solution space accordingly.<n>We propose Vector Optimization with Gaussian Process (VOGP), a probably approximately correct adaptive elimination algorithm that performs black-box vector optimization using Gaussian process bandits.
  arXiv  Detail & Related papers  (2024-12-03T14:47:46Z)
• Faster WIND: Accelerating Iterative Best-of-$N$ Distillation for LLM Alignment [81.84950252537618]
  This paper reveals a unified game-theoretic connection between iterative BOND and self-play alignment.<n>We establish a novel framework, WIN rate Dominance (WIND), with a series of efficient algorithms for regularized win rate dominance optimization.
  arXiv  Detail & Related papers  (2024-10-28T04:47:39Z)
• OptEx: Expediting First-Order Optimization with Approximately Parallelized Iterations [12.696136981847438]
  We introduce first-order optimization expedited with approximately parallelized iterations (OptEx) OptEx is the first framework that enhances the efficiency of FOO by leveraging parallel computing to mitigate its iterative bottleneck. We provide theoretical guarantees for the reliability of our kernelized gradient estimation and the complexity of SGD-based OptEx.
  arXiv  Detail & Related papers  (2024-02-18T02:19:02Z)
• Sample-Efficient "Clustering and Conquer" Procedures for Parallel Large-Scale Ranking and Selection [3.913403111891027]
  We modify the commonly used "divide and conquer" framework in parallel computing by adding a correlation-based clustering step.<n>This seemingly simple modification yields an $mathcalO(p)$ reduction in sample complexity for a widely used class of sample-optimal R&S procedures.<n>In large-scale AI applications such as neural architecture search, our methods demonstrate superior performance.
  arXiv  Detail & Related papers  (2024-02-03T15:56:03Z)
• Large-Scale Sequential Learning for Recommender and Engineering Systems [91.3755431537592]
  In this thesis, we focus on the design of an automatic algorithms that provide personalized ranking by adapting to the current conditions. For the former, we propose novel algorithm called SAROS that take into account both kinds of feedback for learning over the sequence of interactions. The proposed idea of taking into account the neighbour lines shows statistically significant results in comparison with the initial approach for faults detection in power grid.
  arXiv  Detail & Related papers  (2022-05-13T21:09:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.