Divergence-Suppressing Couplings for Rectified Flow

Paper 2 addresses the critical bottleneck of KV cache memory explosion in Large Language Models for long-context tasks. By dynamically generating Soft Tokens and preserving evicted information, it offers a highly timely and practical solution with broad, immediate real-world applications across NLP and industry. While Paper 1 provides a solid methodological improvement for Rectified Flow generative models, Paper 2's focus on LLM efficiency gives it a higher potential for widespread adoption and immediate scientific impact.

Paper 2 likely has higher impact: it introduces a principled, broadly applicable modification to Rectified Flow grounded in field properties (divergence), with clear computational advantages (offline correction, no deployment overhead) and demonstrated gains on synthetic and image-generation tasks. This targets a timely core generative-modeling paradigm and could influence diffusion/flow training practices across domains. Paper 1 is valuable infrastructure (a large benchmark + metrics) but is narrower (programmatic video/code evaluation) and its impact depends on community adoption and alignment with emerging generation paradigms.

Paper 2 addresses a critical and highly timely issue—privacy and security in LLM agents. By introducing a comprehensive benchmark for privacy-utility trade-offs, it provides a crucial tool for AI safety that will likely drive widespread future research and model improvements. While Paper 1 offers a valuable algorithmic enhancement for generative models, Paper 2's focus on LLM privacy has broader real-world applications, cross-disciplinary relevance, and addresses urgent security vulnerabilities in widely deployed systems.

Paper 2 addresses a fundamental issue in Rectified Flow—a widely adopted generative modeling framework—by identifying trajectory entanglement linked to divergence in velocity fields and proposing a principled, computationally efficient correction. This has broad applicability across generative modeling (images, video, etc.) and offers a clean theoretical insight with practical gains at no deployment cost. Paper 1 contributes a useful Bayesian optimization framework for prompt engineering, but operates in a narrower, more applied niche (system prompt tuning) with less fundamental methodological novelty and more limited cross-field impact.

Paper 2 targets a timely, high-visibility area (diffusion/flow-based generative modeling) and proposes a broadly applicable, deployment-neutral correction that can improve sample quality without increasing inference cost—strong real-world relevance and cross-use in generative modeling. The divergence-based diagnosis is conceptually clean and may generalize to other ODE-based models. Paper 1 is novel in formalizing trust calibration as preference learning, but its impact may be narrower (human-in-the-loop agent governance) and leans on established GP classification/PBO machinery with less clear empirical validation from the abstract.

Paper 1 bridges a critical gap by transitioning machine learning applications from controlled lab settings to noisy, real-world educational environments for a highly vulnerable population. Its capacity to predict challenging behaviors 10 minutes in advance offers profound, immediate real-world clinical and societal applications. While Paper 2 presents a solid technical refinement for generative models, Paper 1 demonstrates higher cross-disciplinary impact, translating foundation models into actionable interventions that directly improve human safety and quality of life.

Paper 2 addresses a fundamental issue in Rectified Flow, a widely-used generative modeling framework, with a theoretically grounded and practical solution (divergence-suppressing couplings). It identifies a specific failure mode (trajectory entanglement linked to divergence) and proposes an elegant fix with no additional deployment cost. This has immediate applicability to the large and active generative modeling community. Paper 1, while useful, introduces a benchmark for a relatively niche subproblem (skill generation for LLM agents) that is still emerging. Benchmarks typically have lower impact than methodological innovations unless they become widely adopted standards.

AGPO addresses a fundamental limitation of RLVR methods—reasoning boundary shrinkage—with a novel asymmetric reinforcement strategy that has both theoretical insight and demonstrated practical impact. It shows state-of-the-art results across five math benchmarks and includes large-scale industrial deployment at JD for search ads relevance, demonstrating real-world applicability. Paper 1 offers a useful but incremental improvement to Rectified Flow via divergence suppression, which is narrower in scope and impact. Paper 2's broader relevance to the rapidly growing LLM reasoning field and its dual academic/industrial validation give it higher potential impact.

Paper 1 addresses a fundamental and broadly relevant problem—reproducibility of LLM-assisted scientific analysis—with a practical, deployable solution (typed mediation) validated over six months of real-world use across multiple instruments. It bridges AI and experimental science with clear real-world applications, reducing analysis from weeks to minutes while guaranteeing deterministic outputs. Paper 2 offers a useful but incremental improvement to Rectified Flow training, a narrower contribution within generative modeling. Paper 1's cross-disciplinary impact, practical deployment evidence, and relevance to the reproducibility crisis give it higher potential impact.

Paper 2 proposes a concrete, technically novel modification to Rectified Flow (divergence-suppressing couplings) with clear empirical gains and no added deployment cost, making it readily adoptable in generative modeling pipelines. Its methodological contribution is specific and testable, with direct real-world applications (image generation) and likely downstream use across diffusion/flow-based generative models. Paper 1 offers an important meta-scientific critique and audit methodology for benchmarks, potentially influential in AI evaluation discourse, but its impact is more indirect and harder to operationalize broadly compared to Paper 2’s immediately deployable algorithmic advance.

Paper 1 (SORT) addresses a fundamental and widely-recognized limitation of GRPO-style reinforcement learning for reasoning—the inability to learn from hard prompts where all rollouts fail. This is a critical bottleneck in the rapidly growing field of LLM reasoning. The method is practical, broadly applicable across backbones and benchmarks, and particularly impactful for weaker models (democratizing strong reasoning). Paper 2 offers a useful but more incremental improvement to Rectified Flow via divergence suppression. While technically sound, its scope is narrower (generative modeling) and the gains are more modest. The timeliness and breadth of Paper 1's contribution to LLM reasoning give it higher impact potential.

Paper 1 addresses a critical bottleneck in Rectified Flows, a highly influential and rapidly growing class of generative models. By providing a practical, zero-deployment-cost solution to trajectory entanglement, it offers immediate, broad benefits to high-impact applications like image generation. While Paper 2 offers strong theoretical advancements in MARL, the explosive growth, broader applicability, and rapid adoption of improvements in generative AI give Paper 1 a higher potential for widespread scientific and real-world impact.

Paper 1 addresses a well-defined and practically important problem (multivariate time series anomaly detection) with a novel adversarial learning paradigm that tackles spatial over-generalization. It introduces both a new method and a benchmark with channel-wise annotations, filling a significant evaluation gap. Paper 2 offers a useful but incremental improvement to Rectified Flow via divergence suppression. While technically sound, it is a refinement of an existing method with narrower scope. Paper 1's broader applicability across industrial monitoring, IoT, and cybersecurity, combined with the new benchmark contribution, gives it higher potential impact.

Paper 1 introduces a fundamental algorithmic improvement to Rectified Flow, a highly influential generative modeling framework. By addressing trajectory entanglement mathematically, its contributions can broadly impact various domains using generative models. In contrast, Paper 2 offers a valuable but more narrowly focused empirical study on LLM agent design in a specific POMDP environment. Theoretical and algorithmic advancements in foundational generative models typically yield broader, more long-lasting scientific impact across multiple disciplines.

Paper 2 likely has higher scientific impact: it targets a core issue in rectified flow/diffusion-style generative modeling (trajectory entanglement) with a principled, broadly applicable correction (divergence suppression) that preserves deployment cost. The idea connects to fundamental properties of velocity fields and could influence training/inference procedures across flow-matching and generative modeling work. Paper 1 is timely and useful for GUI agents, but is more application-specific and incremental (training-free region search + MCTS planning) with narrower cross-field reach.

Paper 1 likely has higher impact: it targets a widely used, timely practice (test-time scaling via parallel reasoning/self-verification) and proposes an inference-only method that directly reduces verifier-token cost while improving benchmark performance across multiple models and diverse math/code tasks. Its adaptive compute allocation and diagnostic for suitability are broadly applicable to LLM deployment, offering immediate real-world value. Paper 2 is a more specialized improvement to Rectified Flow; while methodologically interesting and potentially impactful for generative modeling, evidence is narrower (synthetics + image generation) and the application scope is more contained.

Paper 1 introduces a foundational algorithmic improvement to Rectified Flow, a prominent class of generative models. Enhancing trajectory straightness can broadly benefit numerous domains utilizing generative AI, from image synthesis to scientific modeling. Paper 2, while presenting an efficient and practical framework for Emotion Recognition in Conversation, focuses on a specific application within affective computing. Consequently, Paper 1's fundamental methodological contribution is likely to have a broader and deeper scientific impact across multiple fields.