Search AI news

Adaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement Learning

arXiv:2611.08864v1 Announce Type: new Abstract: We present the Adaptive Scheduling Pipeline (ASP), an offline rollout scheduler for multi-instance asynchronous reinforcement learning (RL) of large language models (LLMs). ASP jointly reduces three sources of rollout underutilization - padding bubbles from intra-minibatch length heterogeneity, skewness bubbles from long-tail outputs, and sorting bubbles from fluctuations in training-queue occupancy induced by length-sorted dispatch - through a single offline preprocessing pass. This enables efficient multi-instance rollout that overcomes the early scaling plateau of single-instance inference while smoothing the arrival of generated samples at the trainer. We integrate the scheduler into an asynchronous architecture that decouples rollout from training while keeping policy lag observable and bounded. Complete ASP improves end-to-end throughput by up to 3.22 times and reduces iteration time by up to 76.8% relative to MindSpeed-RL, while reducing iteration time by up to 62.5% relative to verl. Within ASP, scheduling reduces mean iteration time by 10.9%. ASP also scales more robustly with additional generation resources than these baselines and shows no visible convergence degradation in the measured bounded-lag regime. ASP has since been used in an internal telecom-domain RL post-training workflow, reducing the reported training period from more than three months to about two months and producing a domain-enhanced model for knowledge answering and tool calling.

Dual-Stream LLM-Guided Reward Design for Reject-Inference-Aware Reinforcement Learning in E-Commerce Payment Fraud Detection

arXiv:2611.08706v1 Announce Type: new Abstract: Production fraud-detection models learn from a censored slice of their own traffic: transactions a legacy system blocked never receive fraud labels, so models train on a biased, approved-only distribution. We extend LLM-guided reward design, the framework of Qu et al. (2025) in which an LLM evolves the reward of a two-step REINFORCE fraud-blocking policy, to attack this reject-inference problem inside the reward. A single LLM designs a dual-stream reward: stream 1 applies the standard dollar-weighted precision objective to labeled transactions; stream 2 scores legacy-blocked, unlabeled transactions by contrasting the newest gradient-boosted risk signal against the legacy scores that drove the original block. Rather than prescribe which signals matter, we expose a superset of evidence and let the LLM decide for itself which to use, guided by evaluation feedback alone; because it writes the reward as code, that choice is directly auditable. On a held-out four-month out-of-time test, the dual-stream agent beats the prior labeled-only agent at every target recall, by +4.7, +2.4 and +3.0 points of dollar-weighted precision at 80%, 85% and 90% recall, and the production baseline by +6.8 points on average. The resulting policy is architecturally identical to that of Qu et al. (2025) and adds no inference cost.

Guardrail Erosion in Domain Fine-Tuned Large Language Models: A Paired Audit of Knowledge-Centric Abstention in Biomedical LLMs

arXiv:2611.08780v1 Announce Type: new Abstract: The number of domain fine-tuned models, particularly in the biomedical domain, is rapidly expanding. Fine-tuning promises high-performing models at a fraction of the cost of training a base model, but it can also produce unintended behavioral changes. Multiple studies have shown that safety guardrails can erode with fine-tuning, yet large-scale studies across model architectures, data mixtures, and fine-tuning methodologies remain limited. Where prior work in this space has studied value-centric refusal, this work focuses on knowledge-centric abstention. We audit 25 biomedical domain fine-tuned large language models against 11 paired base checkpoints on MedAbstain and ClinDet-Bench to evaluate the impact of fine-tuning on knowledge-centric abstention. We find that defined appropriate abstention measured by these benchmarks, already low in base models (median 8.3% on MedAbstain), falls further after fine-tuning (median 4.2%, decreasing in 21 of 25 pairs on MedAbstain), and that the magnitude and direction of change are heterogeneous across pairs. These results motivate fine-tuning methods that improve domain-specific capability while preserving appropriate caution.

Augmentation with Reinforcement Transformation for Low-textured Scientific Sketch Image Retrieval

arXiv:2611.05408v1 Announce Type: new Abstract: Low-textured sketch images are widely used in scientific applications, including wildlife conservation, archaeological illustration, and forensic composites. Learning from and retrieving such images is challenging because they lack detailed surface textures and visual variations commonly exploited by modern deep learning models. This paper proposes Augmentation with Reinforcement Transformation (ART) to fine-tune pretrained models to learn effective representations for accurate sketch image retrieval. Our key observation is that sketches are particularly sensitive to image transformations, such as rotation, flipping, zooming, and stretching. While some transformations preserve their structural integrity, others may introduce distortions that hinder representation learning. Instead of randomly choosing transformations to construct fine-tuning views, ART formulates augmentation selection as a multi-armed bandit (MAB) problem, where transformations are treated as arms and their effectiveness is continuously evaluated through rewards. This enables ART to adaptively construct reward-guided view banks before self-supervised fine-tuning; the subsequent representation loss does not feed back to ART. On Florida manatee sketches provided by Mote Marine Laboratory, ART achieves 71.40% average Top-1 accuracy for transformed-scar retrieval, compared with 68.83% for random augmentation selection, and 55.84% for location-aware sketch retrieval, compared with 18.18% for frozen CLIP.

Structured Priors: Towards an algebra of component priors for prior-data fitted networks

arXiv:2611.04056v1 Announce Type: new Abstract: Prior-data fitted networks are typically trained using synthetic data generated by a single, manually constructed generator. In cases where such a generator combines multiple components, e.g., in TabPFN, Mitra, or TabICLv2, this is achieved through a mixture using manually specified weights, the impact of which on the network only becomes apparent after pre-training. Instead of such a hand-built generator, this work proposes a structured prior: a small set of named component priors, linked by operators whose algebraic properties are formally proven and whose influence on the trained network has been validated. Within a measure-theoretic framework, the operators of mixture, convolution, amplitude scaling, local transform, and projection are defined. Their governing laws are proven. These operators are implemented in the macl2l software package. Experiments demonstrate that the implemented generators reproduce the identities within the resolution of a permutation test. Furthermore, it is shown that trained networks obey the distributive law; that a mixed prior composed of small components performs as well as or better than any individual component in seven out of eight open regression benchmarks; and that on two of these tables a structured prior reaches, with a ninth of the data budget, an error that a generic random-function prior does not reach even with the full budget. A theorem on model averaging demonstrates that networks trained on individual components constitute the Bayes predictor for the mixture, provided that the component networks are themselves the Bayes predictors for their respective priors. However, the pre-registered test of this averaged predictor with trained component networks failed, because the networks identify the generating component too rarely. The paper proposes a research program focused on the algebra of structured priors for prior-data fitted networks.

Geometric Refinement and Distributional Alignment for Sim2Real Roadside LiDAR Detection

arXiv:2611.03936v1 Announce Type: new Abstract: We describe our approach to the LUMPI track of the UrbanTwin Sim2Real LiDAR Challenge, in which a 3D object detector must be trained exclusively on synthetic data and evaluated on unseen real roadside LiDAR, while the synthetic point clouds are simultaneously scored for distributional realism. Our final entry scores 0.4793 combined, from a detection mAP of 0.1333 and a realism score of 0.9126. Our contribution is not a new architecture. Working from a PointPillars baseline, we obtained most of our gains from three sources. First, a diagnosis that small object failure in roadside LiDAR is a localisation problem rather than a detection problem, which led to a parameter free geometric refinement worth 86 percent relative on pedestrian AP. Second, an audit of our own post processing which revealed that a shipped rule was destroying valid predictions, the repair of which produced our single largest detection gain at 335 percent relative on the Van class. Third, treating the realism metrics analytically rather than as a black box, which showed that one of the four components is a closed form function of nine summary statistics. We also report an experimental methodology that we believe is the more transferable result. Because mean AP decomposes into independent per class terms, a single submission can carry several mutually controlled experiments, and a submission assembled entirely from previously measured components has a score that is known before it is sent. We predicted our final score to four decimal places prior to submission. We report negative results at equal length, and give a full account of data provenance and compliance in Section 8, including a full account of every mechanism contributing to the submitted score.

Supervised Fine-Tuning for Reasoning Large Language Models: A Survey

arXiv:2610.37681v1 Announce Type: new Abstract: Large reasoning models have recently demonstrated remarkable capabilities in solving complex tasks, where supervised fine-tuning (SFT) on long chain-of-thought data serves as a crucial foundation for eliciting and enhancing their reasoning abilities. Despite rapid progress in both improving and analyzing reasoning-oriented SFT, the field still lacks a systematic survey that consolidates its fast-growing literature. To fill this gap, we present a comprehensive review of recent advancements in reasoning SFT, examining the literature through the dual lenses of methodological design and analytical investigation. First, we review methodological improvements across the SFT pipeline and categorize them into data-centric approaches and algorithm-cent

OpenESM: Energy-Steered Models Are Scalable Foundation Learners

arXiv:2610.33005v1 Announce Type: new Abstract: Energy-based models have shown promising properties for domain-specific tasks but remain challenging in modeling general-purpose scenarios. To bridge this gap, we design energy-steered models (ESMs) for foundational language modeling. ESMs integrate energy steering and token generation within a single architecture to leverage the energy-guided search to refine token generation. Specifically, ESMs first optimize a latent state over the energy landscape, then leverages the refined latent state for token generation. By performing such energy-guided pre-training on large-scale corpora, ESMs learn generally effective energy landscape where favorable tokens correspond to lower energy values. To explore the scaling behavior of ESMs, we pre-train ESM

ReMedI: Inference-Time Refinement and Self-Distillation for Controllable Molecular Optimization

arXiv:2610.32103v1 Announce Type: new Abstract: In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to bet

Searching with Sparse Histories: Cross-Task Architecture Search from Incomplete Search Logs and Zero-Cost Proxies

arXiv:2610.32179v1 Announce Type: new Abstract: Neural architecture search (NAS) is expensive, but cross-task transfer can reduce costs by reusing knowledge from previous tasks. However, existing methods struggle with sparse search histories and task-specific transfer failures. We propose a method, a risk-robust strategy that combines sparse source logs with zero-cost proxies through champion-based initialization and joint pairwise-ranker pretraining. On TransNAS-Bench-101, the proposal mitigates severe worst-case failures, reducing worst-10% ranks from 290 to 43 for Micro autoencoding and from 465 to 45 for Macro room layout. Across 13 cell-space targets, it outperforms a proxy-only ranker on 10 targets in both mean and worst-case performance, while maintaining competitive average perform

POS-Signature Demonstration Retrieval for In-Context Dependency Parsing

arXiv:2610.31966v1 Announce Type: new Abstract: Demonstrations for in-context learning are often retrieved by semantic similarity, which may be a poor proxy for relevance in structured prediction. We study demonstration selection for Universal Dependencies parsing with a POS-signature policy that represents sentences with weighted part-of-speech statistics and retrieves three length-compatible, gold-parsed examples from the labeled training split. We evaluate two open-weight language models without fine-tuning on all 2,077 sentences of the UD English Web Treebank test set, comparing the policy with shot- and format-matched fixed and semantic three-shot baselines, as well as zero-shot, chain-of-thought, and critique-refinement systems. POS-signature retrieval yields the highest labeled atta

KnitTrans: Coordinated Trajectory Transfer for Effective and Efficient LLM Reinforcement Learning

arXiv:2610.32389v1 Announce Type: new Abstract: Trajectory transfer from rollout to training affects both learning quality and system efficiency in large language model reinforcement learning (RL). Enforcing batch-wide training requirements can delay the admission of completed trajectories, resulting in higher architectural staleness: the additional parameter-version gap from disaggregated pipelining relative to colocated execution. Beyond admission, misalignment between data access and training consumption causes redundant movement and unnecessary waiting when transferring large trajectory payloads. We present KnitTrans, a system that coordinates the rollout-to-training path through three designs. First, constraint-preserving mini-batch admission allows flexibility in early mini-batch com

Don't Forget What You Just Learned: Self-Directed Repair Turns Agent Mistakes Into Lasting Lessons

arXiv:2610.31846v1 Announce Type: new Abstract: Intelligent agents should learn from their mistakes so that experience leads to continuous improvement. Yet today's agents recover from a mistake only temporarily: in-context reflections are easily forgotten across sessions and compete for limited context, while full-parameter fine-tuning is expensive and can interfere with existing capabilities. We introduce Self-Direction, a mechanism for agents to learn from mistakes that combines the benefits of temporary in-context retries and policy updates. When an agent fails a task, it reflects on its own trajectory in order to extract a lesson and immediately distills it into a lightweight low-rank adapter. By keeping the original weights frozen, Self-Direction prevents adapter training from overwri

Spatial Heuristic Strategy based Diffusion Reinforcement Learning for Online 3D Bin Packing

arXiv:2610.32241v1 Announce Type: new Abstract: Traditional heuristic methods have accumulated mature experience in core objectives, such as space utilization and layout compactness of 3D bin packing, relying on artificially designed rule-based logic. However, they exhibit obvious limitations in generalization ability and adaptability to dynamic scenarios. Diffusion reinforcement learning, which integrates the spatial search capability of reinforcement learning and the data distribution modeling capability of diffusion learning, has achieved certain progress in action distribution modeling and decision compliance constraints. Nevertheless, there remain issues including insufficient perception of the spatial topological structure in 3D bin packing and limited layout optimization capability.

The study proposes a self-assessment reinforcement learning framework, SARL, which enhances the search state-awareness capability of retrieval-augmented large models.

On October 24, 2026, arXiv published the paper arXiv:2610.30435v1, introducing the Self-Assessment Reinforcement Learning (SARL) framework, which aims to address the issue in existing retrieval-augmented large language models where search agents lack the ability to perceive search states.

Why it matters：This study addresses a critical limitation of retrieval-augmented large language models—namely, their tendency to generate hallucinatory reasoning and prematurely terminate searches when they fail to recognize previously acquired or missing information in complex tasks. It proposes an end-to-end reinforcement learning approach that requires no additional evaluation during testing, achieving significant performance improvements on multi-hop reasoning and deep-research benchmarks.

IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems

arXiv:2610.30949v1 Announce Type: new Abstract: We present IQuest-Q1, a sparse Mixture-of-Experts (MoE) agentic foundation model with 320B total parameters (15B activated). IQuest-Q1 is developed through a three-stage process: (1) stable and efficient pre-training, combining Stable SplitNorm, Muon, and muP with a two-stage curriculum whose second stage increases the coding-data mixture, preserving a wide stability margin between safe and optimal learning rates; (2) fine-grained mid-training, expanding the data mixture across domains and formats and progressively extending the context; and (3) multi-scaffold with multi-version post-training, combining SFT, domain-specialized RL, and MOPD across various scaffolds to integrate complementary training signals and strengthen agentic coding capab