paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

Parameter- and Bandwidth-Efficient Edge--cloud Many-to-Many Speech-to-Text Translation

Yexing Du, kaiyuan liu, Youcheng Pan, Bo Yang, Lei Chen, Ming Liu, Bing Qin, Yang Xiang

Multimodal large language models (MLLMs) have demonstrated significant potential for speech-to-text translation (S2TT). However, existing deployment paradigms face critical challenges: pure on-device models suffer from resource constraints, while centralized cloud systems incur bandwidth bottlenecks and privacy risks by transmitting raw voice data. In this paper, we propose Edge--cloud Speech Recognition and Translation (ESRT), a parameter-efficient, bandwidth-efficient, and privacy-aware collaborative Edge--cloud MLLM framework. First, we introduce a multi-task weighted curriculum learning strategy to mitigate catastrophic forgetting, improve multilingual balance, and train parameter-efficient ESRT-1B, ESRT-4B, and ESRT-12B models. Second, we enable bandwidth-efficient Edge--cloud inference by retaining a lightweight speech encoder and adapter on the device and transmitting only a compressed tensor to the cloud. Extensive experiments on FLEURS demonstrate that ESRT models achieve state-of-the-art S2TT performance across 45 languages ($45 \times 44$ directions). Relative to raw audio, ESRT and ESRT-Lite reduce the transmitted tensor size by $5.1\times$ and $10.2\times$, respectively, while keeping raw speech on-device and avoiding its direct exposure to the cloud. The code and models are released to facilitate reproducible, privacy-aware S2TT research.

arXiv

NEURON: A Neuro-symbolic System for Grounded Clinical Explainability

Anuradha Chandrasekaran, Dimitrios Zikos, Mutlu Mete, Alan Pang, Brady D. Lund, Kewei Sha

Clinical AI adoption is hindered by the black-box/grey-box nature of high-performing models, which lack the ontological grounding and narrative transparency required for professional-level explainability. We present NEURON, a neuro-symbolic system designed to enhance both predictive reliability and clinical interpretability. NEURON integrates SNOMED CT ontology-informed structural representations with machine learning models to bridge the gap between raw data and medical nomenclature. To facilitate human-aligned interaction, the system utilizes a Retrieval-Augmented Generation (RAG) grounded Large Language Model (LLM) layer to synthesize SHAP feature attributions and patient-specific clinical notes into coherent, natural-language explanations. Validated on the MIMIC-IV dataset for Acute Heart Failure mortality prediction, NEURON improved the AUC from 0.71-0.74 to 0.74-0.77 and substantially outperformed SHAP-based explanations in human-aligned metrics (0.807 vs. 0.558). Our results demonstrate that NEURON offers a robust, scalable engineering solution for deploying trustworthy, human-centered connected health applications.

arXiv

Bridging Network Fragmentation: A Semantic-Augmented DRL Framework for UAV-aided VANETs

Gaoxiang Cao, Wenke Yuan, Huasen He, Yunpeng Hou, Xiaofeng Jiang, Shuangwu Chen, Jian Yang

Urban Vehicular Ad-Hoc Networks (VANETs) can become fragmented because buildings obstruct wireless links and vehicle mobility continuously changes the network topology. Unmanned Aerial Vehicles (UAVs) can serve as mobile relays, but Deep Reinforcement Learning (DRL)-based deployment often suffers from inefficient exploration because it lacks road-topology guidance. To address this problem, we propose Semantic-Augmented DRL (SA-DRL), which models network fragmentation over the road topology and aligns a pretrained Large Language Model (LLM) to generate a topology-dependent action prior from dynamic traffic states. The resulting Semantic-Augmented PPO (SA-PPO) algorithm combines this prior with the PPO policy through Logit Fusion, guiding exploration toward promising intersections while retaining adaptation through environmental returns. Simulations driven by real-world urban trajectories show that SA-PPO reaches the final converged reward of Vanilla PPO using only 28.6% of its training episodes. It improves the average number of vehicles in connected components and the average connected-component size by 7.9% and 8.7%, respectively, while reducing UAV energy consumption by 21.3%.

arXiv

Projection-based multifidelity linear regression for data-scarce applications

Vignesh Sella, Julie Pham, Karen Willcox, Anirban Chaudhuri

Surrogate modeling for systems with high-dimensional quantities of interest remains challenging, particularly when training data are costly to acquire. This work develops multifidelity methods for multiple-input multiple-output linear regression targeting data-limited applications with high-dimensional outputs. Multifidelity methods integrate many inexpensive low-fidelity model evaluations with limited, costly high-fidelity evaluations. We introduce two projection-based multifidelity linear regression approaches with linear and nonlinear features that leverage principal component basis vectors for dimensionality reduction and combine multifidelity data through: (i) a direct data augmentation using low-fidelity data, and (ii) a data augmentation incorporating explicit linear corrections between low-fidelity and high-fidelity data. The data augmentation approaches combine high-fidelity and low-fidelity data into a unified training set and train the linear regression model through weighted least squares with fidelity-specific weights. We introduce a proximity-based weighting scheme with automatic weight selection strategy through cross-validation. The proposed multifidelity linear regression methods are demonstrated on approximating the surface pressure field of a hypersonic vehicle in flight and the temperature field on an aircraft disc braking system. In an ultra low-data regime of no more than twelve high-fidelity samples, multifidelity linear regression achieves approximately 2%-12% improvement in median accuracy and a higher $R^2$ score relative to single-fidelity methods at comparable computational cost.

arXiv

AtomBridge: Agentic VLA Inference Plugin for Long-Horizon Tasks in Scientific Experiments

Yiwen Pang, Bo Zhou, Changjin Li, Xuanhao Wang, Shengxiang Xu, Deng-Bao Wang, Peng Cheng, Shimin Di, Jingkuan Song, Min-Ling Zhang

Robotic laboratories play a critical role in autonomous scientific discovery by enabling scalable, continuous experimental execution. Recent vision-language-action (VLA) models offer a promising foundation for robotic laboratories. However, scientific experiments typically involve long-horizon tasks composed of multiple atomic tasks. Existing VLA models may fail to perform composed tasks formed by reordering and composing these known atomic actions. This limitation can arise from a skill-chaining gap caused by robot-state mismatch: the terminal robot state of one skill can fall outside the valid initial-state distribution of the next. To address this challenge, we propose AtomBridge, an Agentic VLA Inference Plugin for Long-Horizon Tasks in Scientific Experiments. AtomBridge attaches at inference time to a VLA policy already fine-tuned on atomic tasks, while keeping its weights fixed. At each task boundary, it uses LLM-based transition reasoning and robotic-action code generation to insert transitional actions between consecutive tasks. This plug-and-play design mitigates the skill-chaining gap caused by robot-state mismatch without additional VLA fine-tuning or demonstrations of composed long-horizon sequences. Across scientific manipulation sequences in simulation and a real-world experimental environment, AtomBridge improves execution continuity and per-step atomic-task success. On 8-step composed tasks, AtomBridge improves full-sequence success by 10%~25%.

arXiv

FDA-Opt: Federated Fine-Tuning via Dynamic Update Schedules

Michael Theologitis, Vasilis Samoladas, Antonios Deligiannakis

Federated Learning (FL) enables the utilization of vast, previously inaccessible data sources. At the same time, pre-trained Language Models (LMs) have taken the world by storm and for good reason. They exhibit remarkable emergent abilities and are readily adapted to downstream tasks. This opens one of the most exciting frontiers in FL: fine-tuning LMs. Yet, a persistent challenge in FL is the frequent, rigid communication of parameters -- a problem magnified by the sheer size of these contemporary models. The FedOpt family of algorithms has become the go-to approach for FL, relying on fixed but arbitrary intervals for model exchanges. Recently, the FDA algorithm prescribed a dynamic approach by monitoring the training progress. However, it introduced a hard-to-calibrate parameter and imposed a rigid synchronization scheme. In this work, we address these limitations by proposing the FDA-Opt family of algorithms -- a unified generalization of both FDA and FedOpt. Our experimental evaluation focuses on fine-tuning LMs on downstream NLP tasks and demonstrates that FDA-Opt outperforms FedOpt even when it is configured with hyper-parameters specifically optimized for the latter. In other words, we show that FDA-Opt is a practical, drop-in replacement for FedOpt in modern FL libraries and systems: it requires no additional configuration and delivers superior performance out of the box.

arXiv

The Metacognitive Bottleneck: Japanese Riddles Reveal Fundamental Limits of Machine Insight and Self-Evaluation in Reasoning AI

Masaharu Mizumoto, Dat Nguyen, Zhiheng Han, Xingfu Li, Yo Nakawake, Le Minh Nguyen

Benchmark saturation and training-data contamination increasingly obscure whether reported gains in large language models (LLMs) reflect genuine advances in reasoning or familiarity with recurring patterns in benchmark problems. We introduce the NazoNazo Benchmark, a renewable and extensible evaluation dataset derived from Japanese children's riddles that isolates a specific class of reasoning processes: insight-like representational restructuring and metacognitive evaluation. Rather than modeling reasoning in general, these tasks provide a focused test of failure modes that are difficult to detect in standard benchmarks. We curate 201 riddles and establish a human reference on a 120-item subset (n = 126; mean accuracy 52.9%). The benchmark is fully open, low-cost to refresh, and designed for continual evaluation under reduced contamination risk. We evaluate 38 frontier LLMs (2023-2025) under a strict retrieval-free, zero-shot protocol. On the human-comparison subset, non-reasoning models achieve 7.6% accuracy and reasoning-oriented models reach 17.6%, compared with a human mean of 52.9%, although performance varies substantially across models. Beyond accuracy, qualitative analysis of model-generated thought-logs identifies a distinctive failure mode, which we call verification failure: models generate a correct intermediate candidate but fail to endorse it as their final answer. This dissociation between candidate generation and endorsement reveals a metacognitive bottleneck: across the models with usable thought-logs, verification failures account for between 5% and 39% of a model's incorrect answers. By isolating the gap between generation and verification, this work provides a practical framework for diagnosing reasoning reliability and suggests concrete directions for improvement, including better calibration, structured verification, and stopping mechanisms.

arXiv

LLM-Advisor: An LLM Advisor for Cost-efficient Path Planning across Multiple Terrains

Ling Xiao, Toshihiko Yamasaki

This paper addresses fixed-graph terrain-aware path refinement, in which a global planner is restricted to a predefined route space and may remain optimal within that space while missing lower-cost terrain corridors available in the native-resolution map. We propose LLM-Advisor, an external verification-guided multimodal refinement framework that uses semantic terrain context to propose route alternatives with lower cost than the baseline route. The framework combines a multimodal prompt, map-disjoint Reference Example Augmentation (REA), and deterministic validation with fallback: a proposed route is deployed only when it is feasible and strictly lower in cost than the baseline. To evaluate the method, we introduce MultiTerraPath, a controlled 2,000-map benchmark with Easy and Hard subsets for fixed-graph terrain-cost refinement, and further conduct a semantic-cost transfer evaluation on RUGD. Using coarse-lattice A* with stride r=50 as the fixed baseline planner, we compare LLM-Advisor with direct LLM planning and LLM-A*. LLM-Advisor with GPT-5.5 achieves the highest FIR on MultiTerraPath, improving 52.00% of Easy maps and 45.00% of Hard maps. Deterministic verification and fallback yield 100.00% NDR and Final Validity across all evaluated RUGD scene groups. Across lattice strides from r=10 to r=100, offline oracle analysis shows that coarser graphs create larger coarse--oracle gaps, while LLM-Advisor recovers up to 49.16% and 29.98% of the available gap on Easy and Hard maps when r=100, respectively.

arXiv

From Field Data to Global Food Systems Intelligence: A Semantic Graph Framework for Sustainable Wheat Production

Nirmal Gelal, Aastha Gautam, Soheil Abadifard, Nico Giordano, Moumita Sen Sarma, Sanaz Saki Norouzi, Claudio Dias da Silva Jr, Jean Ribert Francois, Kathleen M. Jagodnik, Katherine Nelson, Terry Griffin, Xiaomao Lin, Stacy Hutchinson, Stephen M. Welch, Kelsey Andersen Onofre, Romulo Lollato, Pascal Hitzler, Hande Küçük McGinty

In response to the growing need for structured, interoperable agricultural data, this paper presents the Sustainable Wheat Production Datahub, a modular, graph-based framework that brings diverse wheat production datasets together into a single, queryable store. Using the Knowledge Acquisition and Representation Methodology (KNARM), with domain experts in the loop, we developed ontologies for nutrient management and disease management, two of the areas that most affect wheat yield and its sustainability, covering practices such as nitrogen fertilization and fungicide-based disease control. The two ontologies are federated, meaning they are maintained as separate but connected modules, joined by a bridging layer of cross-domain links, and together they form the schema of a knowledge graph (KG). We construct this KG by populating the ontologies with data from diverse sources, including field trials, expert knowledge, and environmental descriptors. We validate the ontologies, showing the KG accurately answers practical agronomic questions expressed in SPARQL. We demonstrate that the KG derives facts entailed by the ontologies beyond those explicitly stored, and that a single query can draw across independently sourced datasets. The framework also has practical implications for wheat research and extension programs, since it makes data from different sources easier to find, combine, and reuse. Designed to expand toward the full wheat lifecycle, from farm to table, this work establishes the foundation for a scalable, semantically rich global food systems datahub.

arXiv

Musical Agent Systems: MACAT and MACataRT

Keon Ju M. Lee, Philippe Pasquier

Our research explores the development and application of musical agents, human-in-the-loop generative AI systems designed to support music performance and improvisation within co-creative spaces. We introduce MACAT and MACataRT, two distinct musical agent systems crafted to enhance interactive music-making between human musicians and AI. MACAT is optimized for agent-led performance, employing real-time synthesis and self-listening to shape its output autonomously, while MACataRT provides a flexible environment for collaborative improvisation through audio mosaicing and sequence-based learning. Both systems emphasize training on personalized, small datasets, fostering ethical and transparent AI engagement that respects artistic integrity. This research highlights how interactive, artist-centred generative AI can expand creative possibilities, empowering musicians to explore new forms of artistic expression in real-time, performance-driven and music improvisation contexts.

arXiv

Edge Case Detection in Automated Driving: Methods, Challenges, and Future Directions

Saeed Rahmani, Sabine Rieder, Erwin de Gelder, Marcel Sonntag, Jorge Lorente Mallada, Sytze Kalisvaart, Vahid Hashemi, Bart van Arem, Simeon C. Calvert

Automated vehicles (AVs) promise to enhance transportation safety and efficiency. However, ensuring their reliability in real-world conditions remains challenging, particularly due to rare and unexpected situations known as edge cases. While numerous approaches exist for detecting edge cases, a comprehensive survey reviewing these techniques is lacking. This paper bridges this gap by presenting a hierarchical review and systematic classification of edge case detection and assessment methodologies. Our classification is structured on two levels: first, by AV modules, including perception and trajectory-related (encompassing prediction, planning, and control) subsystems; and second, by underlying methodologies and theories guiding these techniques. Furthermore, we introduce "knowledge-driven" approaches, which complement data-driven methods by leveraging expert insights and domain knowledge to identify cases absent in training datasets. We then examine techniques and metrics for evaluating edge case detection methods, including detection performance, practical deployment (e.g., computational overhead), and domain-specific measures (e.g., crash rates and severity analysis). We conclude by highlighting key challenges for edge case detection, including data availability and quality issues, validation and interpretability limitations, the simulation-to-real gap, and computational constraints. The hierarchical classification and review of methods and assessment techniques in this survey enable modular and targeted testing frameworks by guiding the selection of detection methods for specific AV subsystems while considering methodological principles. It also supports practical testing by facilitating scenario generation in simulation and focused subsystem validation in the real world.

arXiv

Balancing Optimality and Diversity: Human-Centered Decision Making through Generative Curation

Michael Lingzhi Li, Shixiang Zhu

Many decision-support systems recommend actions by optimizing measurable objectives, even when a human decision-maker retains final authority and considers additional criteria that are difficult to specify in advance. We study how an algorithm should curate a small portfolio of quantitatively strong alternatives in such settings. We introduce generative curation, a framework that learns a recommendation policy to maximize the expected desirability of the action ultimately selected by the decision-maker. For policies that generate quantitatively competitive actions, we decompose expected portfolio desirability into quantitative performance and a qualitative curation gain. Under a Gaussian process model of residual desirability, this gain is characterized by the Gaussian width induced by the covariance kernel, yielding a decision-theoretic notion of diversity based on qualitative nonredundancy rather than generic geometric separation. We establish diminishing returns to portfolio size and characterize regimes in which optimal policies are balanced, endpoint-concentrated, or space-filling. We develop neural generative and sequential optimization approaches applicable to continuous and combinatorial action spaces. Controlled synthetic experiments demonstrate substantial regret reductions relative to optimization- and distance-based benchmarks, while an Atlanta police redistricting case study illustrates the framework's applicability to a complex operational planning problem.

arXiv

RecipeNet: A Hierarchical Transformer for Recipe Data

Pin-Yen Huang, Sachin Chhabra, Prasanth Sai Gouripeddi, Abhinav Kumar, Baoxin Li

Recipe data arises in domains such as materials synthesis, pharmaceutical formulation, and industrial manufacturing, where procedures are represented as ordered sequences of steps containing heterogeneous structured fields. Existing tabular learning methods typically flatten this structure into fixed-schema representations, limiting their ability to capture hierarchical field interactions and procedural dependencies. We propose RecipeNet, a hierarchical Transformer architecture that encodes field-level interactions within each step and sequential dependencies across steps through stacked Transformer encoders. Experiments on multiple recipe datasets and tasks demonstrate that RecipeNet consistently outperforms existing tabular models, highlighting the value of hierarchical and sequential modeling for recipe representation learning.

arXiv

Universal Thermodynamic Interatomic Potentials for Crystalline Materials

Juno Nam, Bowen Deng, Xiaochen Du, Luis Barroso-Luque, Benjamin Kurt Miller, Rafael Gómez-Bombarelli

Free energies govern solid-state phase stability, yet computational materials discovery still relies largely on ground-state energies because free energy calculations require ensemble averages. We introduce the thermodynamic interatomic potential (TIP), which extends an interatomic potential from its static energy to a thermodynamically consistent Gibbs free energy model, with thermodynamic responses following from temperature and pressure by automatic differentiation. We implement TIP[UMA] using the universal potential UMA, train it on free energies from quasi-harmonic to molecular dynamics fidelity, and calibrate it to higher-resolution calculations or experiment. From a single evaluation, it returns the equation of state of a crystal and locates phase transitions among competing branches, including dynamically stabilized phases. Fine-tuning extends the model to alloy solubility limits and miscibility gaps. TIP makes the free energy as accessible as the potential energy, opening finite-temperature phase stability to high-throughput discovery.

arXiv

Rollplex: Cross-Phase GPU Spatial Sharing for Vision Language Model Post-Training

Hanfeng Lu, Tianyu Feng, Suyi Li, Yuheng Zhao, Wei Gao, Shaopan Xiong, Ju Huang, Siran Yang, Jiamang Wang, Lin Qu, Wei Wang

Vision-language models (VLMs) enable embodied agents to reason and act from visual observations and language instructions. Reinforcement learning (RL) post-training enhances these capabilities using task feedback, but current on-policy RL runtimes execute rollout, reference scoring, and actor training in strict serial phases. While effective for text-only RL, this phase-granular execution is wasteful for VLMs, where processing dense video inputs and prompt prefixes occupies a large fraction of each phase. Because prefix processing is independent of the generated response, it can be run alongside rollout decoding, which leaves GPU compute capacity underutilized, without breaking synchronous on-policy semantics. We present Rollplex, a runtime that decomposes the reference and training phase and moves the prefix computation into the rollout decode window. Realizing this schedule requires more than concurrent kernel launches: naive colocation of Qwen2.5-VL-32\,B requires roughly 165\,GiB per GPU, while rollout and training prefer different tensor-parallel (TP) degrees and weight layouts. Rollplex addresses these constraints with two mechanisms. Phase-aware memory management controls HBM residency according to producer--consumer lifetimes. Parallelism-aware weight sharing uses the same physical storage for layout-compatible tensors across distinct TP degrees and reconstructs only incompatible tensors, avoiding a complete second actor copy. On 32 H800 GPUs, Rollplex achieves $1.23\times$--$1.30\times$ speedup over serial colocation and $1.57\times$--$2.24\times$ over disaggregation under the same GPU budget, while preserving the synchronous RL update.

arXiv

Generating Benchmark Health Data Using a Tabular Diffusion Transformer

Hao Yan, Lisa Pilgram, Dan Liu, Linglong Kong, Fida Dankar, Khaled El Emam

Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.

arXiv

Approximate Muon with low-rank adapters

Ben Anson, Conor Houghton, Edward Milsom

The Muon optimizer shows clear benefits versus alternatives when pretraining neural networks. However, it is used less frequently for parameter-efficient fine-tuning (PEFT). One potential reason is that the most common PEFT method, LoRA, does not naturally combine with Muon since it is not mathematically possible to orthogonalize the weight update given by a low-rank parameterization. In this paper, we address this issue by approximating the solution to a relaxed Muon objective in the low-rank setting via linearization and then least-squares. We provide an efficient implementation that uses matmul operations only, as opposed to more complex linear algebra decomposition routines. Our method, sMuon (small Muon), performs favourably across SFT and a ReLoRA pretraining experiment. While results are model- and eval-dependent, we find overall that using Muon for low-rank fine-tuning provides moderate performance improvements.

arXiv

LP-NAS: Linear Programming-based Neural Architecture Search

Abhishek Shukla, Ankur Sinha, Faiz Hamid

Neural Architecture Search (NAS) aims to automate neural network architecture design, reducing reliance on human expertise. Among the various NAS methods, differentiable NAS has gained prominence due to its efficiency and accuracy compared to conventional NAS approaches. Since differentiable NAS relaxes the architecture search space into a continuous domain, it is possible to apply principles from continuous optimization to NAS. In this paper, we propose Linear Programming-based NAS (LP-NAS), a mathematical programming-based framework for differentiable NAS that is applicable to a wide range of continuous search spaces. LP-NAS formulates a linear program (LP) using the validation-loss gradient and the training-loss Hessian to compute an architecture update direction that improves generalization while preserving the optimality of the model parameters. By following this LP-derived descent direction, LP-NAS efficiently navigates the architecture search space, leading to faster and more effective architecture optimization. We introduce two computationally efficient variants of LP-NAS, namely S-LP-NAS and R-LP-NAS. Applying LP-NAS to the Differentiable Architecture Search (DARTS) search space results in two algorithmic variants, S-LP-DARTS and R-LP-DARTS. Both variants achieve faster convergence and significantly higher validation performance during the early search iterations than the standard DARTS algorithm. Extensive experiments on CIFAR-10 and CIFAR-100 show that LP-DARTS outperforms standard DARTS in both the architecture search and evaluation phases. Additionally, we compare our approach with several DARTS variants (P-DARTS, PC-DARTS, and STO-DARTS) on the CIFAR-10 dataset and demonstrate its effectiveness. Furthermore, we validate the transferability of the discovered architectures through experiments on the ImageNet dataset.

arXiv

Expected Free Energy-based Informative Path Planning for Robotic Mars Exploration

Ajith Anil Meera, Pablo Lanillos, Wouter Kouw

An autonomous robot efficiently exploring an unknown environment, such as looking for water sources on Mars, faces two simultaneous demands: building an accurate information map while quickly finding the regions of greatest value, and paying for every meter of travel and the cost of every measurement it takes. Classical information-seeking and reward-seeking criteria address only one of these objectives at a time. Here, we propose Expected Free Energy (EFE), the principled action-selection objective from active inference, as a unifying criterion for budgeted robotic informative path planning. Maintaining a Gaussian-process belief over the information field, our agent plans continuous trajectories that minimize expected free energy under hard path-length constraints. The results from multiple realizations show that EFE-based planning yields accurate posterior maps and locates the highest-value regions simultaneously, outperforming information-theoretic baselines under the same settings. In robotic exploration, these unified, easy-to-tune principled information-gathering strategies facilitate autonomous deployment while enforcing efficiency and resource constraints.

arXiv

Designing Compact Neural Architectures via Neuron Gating and Mixed Activation

Abhishek Shukla, Ankur Sinha, Faiz Hamid

Neural Architecture Search (NAS) is naturally formulated as a bilevel optimization problem, where the upper-level optimizes the architecture using validation performance and the lower-level trains network parameters using training loss. However, NAS is computationally expensive due to discrete architectural decisions, exponentially growing search spaces, and the high cost of training candidate architectures. This work develops a general bilevel optimization framework for NAS across diverse architectures, including MLPs, CNNs, RNNs, and Transformers, to identify compact architectures with strong predictive performance. We propose three scalable formulations that replace discrete neuron- and activation-level decisions with continuous relaxations, enabling differentiable optimization over otherwise combinatorial architecture spaces. These formulations give rise to three NAS methods: NAS based on Neuron Gating (NAS-NG), NAS based on Mixed Activation (NAS-MA), and NAS based on Neuron Gating and Mixed Activation (NAS-NGMA). Experiments on MLPs and CNNs using MNIST and CIFAR-10 show that the proposed methods consistently identify compact architectures with competitive or improved predictive performance. On MNIST, NAS-NGMA achieves 98.68% test accuracy with 7.69M MLP parameters, while NAS-NG achieves 99.63% accuracy with only 0.26M CNN parameters. On CIFAR-10, the proposed methods consistently outperform vanilla DARTS. Further experiments demonstrate that NAS-NG can optimize substantially over-parameterized and literature-optimal architectures, improving accuracy while reducing parameters. These results establish relaxed bilevel optimization as a scalable alternative to discrete NAS and provide a general framework for efficient neuron- and activation-level architecture optimization.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.