42 Search Results for "Tan, Cheng"

Document
DOI: 10.4230/LIPIcs.FORC.2026.7
When to Ask a Question: Understanding Communication Strategies in Generative AI Tools

Authors: Charlotte Park, Kate Donahue, and Manish Raghavan

Published in: LIPIcs, Volume 368, 7th Symposium on Foundations of Responsible Computing (FORC 2026)

Abstract
Generative AI models differ from traditional machine learning tools in that they allow users to provide as much or as little information as they choose in their inputs. This flexibility often leads users to omit certain details, relying on the models to infer and fill in under-specified information based on distributional knowledge of user preferences. Such inferences may privilege majority viewpoints and disadvantage users with atypical preferences, raising concerns about fairness. Unlike more traditional recommender systems, LLMs can explicitly solicit more information from users through natural language. However, while directly eliciting user preferences could increase personalization and mitigate inequality, excessive querying places a burden on users who value efficiency. We develop a stylized model of user-LLM interaction and develop an objective that captures tradeoff between user burden and preference representation. Building on the observation that individual preferences are often correlated, we analyze how AI systems should balance inference and elicitation, characterizing the optimal amount of information to solicit before content generation. Ultimately, we show that information elicitation can mitigate the systematic biases of preference inference, enabling the design of generative tools that better incorporate diverse user perspectives while maintaining efficiency. We complement this theoretical analysis with an empirical evaluation illustrating the model’s predictions and exploring their practical implications.
Cite as

Charlotte Park, Kate Donahue, and Manish Raghavan. When to Ask a Question: Understanding Communication Strategies in Generative AI Tools. In 7th Symposium on Foundations of Responsible Computing (FORC 2026). Leibniz International Proceedings in Informatics (LIPIcs), Volume 368, pp. 7:1-7:25, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{park_et_al:LIPIcs.FORC.2026.7,
  author =	{Park, Charlotte and Donahue, Kate and Raghavan, Manish},
  title =	{{When to Ask a Question: Understanding Communication Strategies in Generative AI Tools}},
  booktitle =	{7th Symposium on Foundations of Responsible Computing (FORC 2026)},
  pages =	{7:1--7:25},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-419-2},
  ISSN =	{1868-8969},
  year =	{2026},
  volume =	{368},
  editor =	{Lin, Huijia (Rachel)},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.FORC.2026.7},
  URN =		{urn:nbn:de:0030-drops-259782},
  doi =		{10.4230/LIPIcs.FORC.2026.7},
  annote =	{Keywords: human-AI interaction, user modeling, personalization}
}
Document
DOI: 10.4230/LIPIcs.SoCG.2026.78
Mapping Chemical Space: Topological Data Analysis of Chemical Latent Space with Mapper

Authors: Dhruv Meduri, Chuan-Shen Hu, Cong Shen, Kelin Xia, and Bei Wang

Published in: LIPIcs, Volume 367, 42nd International Symposium on Computational Geometry (SoCG 2026)

Abstract
The vast chemical space, encompassing virtually innumerable molecules and materials, presents both immense opportunities and significant challenges. The design and discovery of novel drugs and functional materials may be viewed as a search within this space; however, the sheer scale of potential candidates renders exhaustive exploration infeasible. To address this, we introduce Chemical Mapper, a framework that integrates topological data analysis with deep learning to enable the visual exploration and analysis of chemical latent spaces. At its core, Chemical Mapper employs mapper, a widely used tool in topological data analysis, to investigate the organizational principles of chemical latent spaces defined by molecular representations learned by geometric deep learning models. In doing so, Chemical Mapper not only highlights groups of molecular representations but also uncovers the relationships among them through linkages and branching structures. Our results show that Chemical Mapper reveals intrinsic patterns associated with molecular scaffolds, functional groups, and chemical properties, as well as the structural and functional evolutions of the molecules.
Cite as

Dhruv Meduri, Chuan-Shen Hu, Cong Shen, Kelin Xia, and Bei Wang. Mapping Chemical Space: Topological Data Analysis of Chemical Latent Space with Mapper. In 42nd International Symposium on Computational Geometry (SoCG 2026). Leibniz International Proceedings in Informatics (LIPIcs), Volume 367, pp. 78:1-78:20, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{meduri_et_al:LIPIcs.SoCG.2026.78,
  author =	{Meduri, Dhruv and Hu, Chuan-Shen and Shen, Cong and Xia, Kelin and Wang, Bei},
  title =	{{Mapping Chemical Space: Topological Data Analysis of Chemical Latent Space with Mapper}},
  booktitle =	{42nd International Symposium on Computational Geometry (SoCG 2026)},
  pages =	{78:1--78:20},
  series =	{Leibniz International Proceedings in Informatics (LIPIcs)},
  ISBN =	{978-3-95977-418-5},
  ISSN =	{1868-8969},
  year =	{2026},
  volume =	{367},
  editor =	{Ahn, Hee-Kap and Hoffmann, Michael and Nayyeri, Amir},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.SoCG.2026.78},
  URN =		{urn:nbn:de:0030-drops-258854},
  doi =		{10.4230/LIPIcs.SoCG.2026.78},
  annote =	{Keywords: Practice of computational topology, topological data analysis, applications in chemistry, mapper algorithm, high-dimensional data analysis, chemical spaces, geometric deep learning, latent space geometry}
}
Document
Research
DOI: 10.4230/TGDK.4.1.4
Native Provenance Computation for Federated and Non-Federated SPARQL Queries

Authors: Zubaria Asma, Daniel Hernández, Luis Galárraga, Giorgos Flouris, Irini Fundulaki, and Katja Hose

Published in: TGDK, Volume 4, Issue 1 (2026). Transactions on Graph Data and Knowledge, Volume 4, Issue 1

Abstract
The popularity of knowledge graphs (KGs) owes credit to their flexible data model, which is suitable for data integration from multiple sources. Several KG-based applications, such as trust assessment, view maintenance, or data valuation on dynamic data, rely on the ability to compute provenance explanations for query results. This need becomes more urgent in federated query processing systems, which allow the online consumption of heterogeneous and decentralized Web data. However, the problem of computing and interacting with provenance has received little attention, especially in the federated setting. On those grounds, this paper introduces the NPCS (Native Provenance Computation for SPARQL) approach, and its federated variant Fed-NPCS, that compute provenance for SPARQL query results. Both approaches build upon spm-semirings to annotate the results of monotonic and non-monotonic SPARQL queries with their provenance. Due to their reliance on query rewriting techniques, the approaches are directly applicable to already deployed SPARQL engines and federations using different reification schemes, including RDF-star. Our experimental evaluation shows that our novel query rewriting approach brings significant run-time improvements w.r.t. the state-of-the-art across both centralized and federated settings. In centralized settings, our tests on two popular SPARQL engines (GraphDB and Stardog) reveal substantial runtime gains over existing query rewriting solutions, enabling scalability to RDF graphs with billions of triples. In federated settings, our experiments on the FedShop benchmark with GraphDB show the viability of Fed-NPCS for federations with up to 200 sources.
Cite as

Zubaria Asma, Daniel Hernández, Luis Galárraga, Giorgos Flouris, Irini Fundulaki, and Katja Hose. Native Provenance Computation for Federated and Non-Federated SPARQL Queries. In Transactions on Graph Data and Knowledge (TGDK), Volume 4, Issue 1, pp. 4:1-4:43, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@Article{asma_et_al:TGDK.4.1.4,
  author =	{Asma, Zubaria and Hern\'{a}ndez, Daniel and Gal\'{a}rraga, Luis and Flouris, Giorgos and Fundulaki, Irini and Hose, Katja},
  title =	{{Native Provenance Computation for Federated and Non-Federated SPARQL Queries}},
  journal =	{Transactions on Graph Data and Knowledge},
  pages =	{4:1--4:43},
  ISSN =	{2942-7517},
  year =	{2026},
  volume =	{4},
  number =	{1},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/TGDK.4.1.4},
  URN =		{urn:nbn:de:0030-drops-259642},
  doi =		{10.4230/TGDK.4.1.4},
  annote =	{Keywords: native provenance computation, federated SPARQL queries, data provenance, NPCS, Fed-NPCS}
}
Document
DOI: 10.4230/OASIcs.PARMA-DITAM.2026.8
Performance Modeling & Mapping of LLM Inference on Heterogeneous Vectorized CGRAs

Authors: Dionysios Kefallinos, Georgios Alexandris, Alexis Maras, Panagiotis Chaidos, Manil Dev Gomony, Henk Corporaal, Dimitrios Soudris, and Sotirios Xydis

Published in: OASIcs, Volume 141, 17th Workshop on Parallel Programming and Run-Time Management Techniques for Many-Core Architectures and 15th Workshop on Design Tools and Architectures for Multicore Embedded Computing Platforms (PARMA-DITAM 2026)

Abstract
Since the emergence of transformer-based models, the computational demands for Large Language Model (LLM) inference have been increasing exponentially, primarily due to their compounding parameter sizes, their structural complexity, and the use of non-linear functions. This tendency leads to the necessity of deploying them on low-power edge devices and DNN accelerators, to fuel next-generation agentic AI systems. Coarse-Grained Reconfigurable Architectures (CGRAs) have proven to be a compelling paradigm for edge acceleration, combining the programmability of general-purpose platforms with the high performance and energy efficiency associated with ASICs. In this work, we introduce an end-to-end performance modeling and mapping framework for LLM inference on heterogeneous CGRAs. Our methodology enables rapid exploration of the micro-architectural design space parameters, i.e., the number of processing elements, vector sizes, and memory configurations, by providing an accurate, explainable, and analytical CGRA performance modeling methodology, with an average cycle error of 0.9%. Architecturally, we build upon R-Blocks, a heterogeneous CGRA platform, and extend it to support floating-point arithmetic operations as well as a full-stack compilation and mapping flow for both full (FP32) and quantized (INT8) Llama2 models. The proposed methodology, evaluated on a 22nm technology node, achieves superior peak performance per Watt compared to related works such as REVAMP and CFEACT (1.8× and 2.8× respectively).
Cite as

Dionysios Kefallinos, Georgios Alexandris, Alexis Maras, Panagiotis Chaidos, Manil Dev Gomony, Henk Corporaal, Dimitrios Soudris, and Sotirios Xydis. Performance Modeling & Mapping of LLM Inference on Heterogeneous Vectorized CGRAs. In 17th Workshop on Parallel Programming and Run-Time Management Techniques for Many-Core Architectures and 15th Workshop on Design Tools and Architectures for Multicore Embedded Computing Platforms (PARMA-DITAM 2026). Open Access Series in Informatics (OASIcs), Volume 141, pp. 8:1-8:14, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{kefallinos_et_al:OASIcs.PARMA-DITAM.2026.8,
  author =	{Kefallinos, Dionysios and Alexandris, Georgios and Maras, Alexis and Chaidos, Panagiotis and Gomony, Manil Dev and Corporaal, Henk and Soudris, Dimitrios and Xydis, Sotirios},
  title =	{{Performance Modeling \& Mapping of LLM Inference on Heterogeneous Vectorized CGRAs}},
  booktitle =	{17th Workshop on Parallel Programming and Run-Time Management Techniques for Many-Core Architectures and 15th Workshop on Design Tools and Architectures for Multicore Embedded Computing Platforms (PARMA-DITAM 2026)},
  pages =	{8:1--8:14},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-416-1},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{141},
  editor =	{Baroffio, Davide and Busia, Paola and Denisov, Lev and Shukla, Nitin},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.PARMA-DITAM.2026.8},
  URN =		{urn:nbn:de:0030-drops-256752},
  doi =		{10.4230/OASIcs.PARMA-DITAM.2026.8},
  annote =	{Keywords: Edge AI, LLM, CGRA, Heterogeneous Architectures, Performance Modeling, Hardware Acceleration, Low Power Computing}
}
Document
Research
DOI: 10.4230/TGDK.4.1.1
On the Computational Cost of Knowledge Graph Embeddings

Authors: Victor Charpenay, Mansour Zoubeirou A Mayaki, and Antoine Zimmermann

Published in: TGDK, Volume 4, Issue 1 (2026). Transactions on Graph Data and Knowledge, Volume 4, Issue 1

Abstract
Over a decade, numerous Knowledge Graph Embedding (KGE) models have been designed and evaluated on reference datasets, always with increasing performance. In this paper, we re-evaluate these models with respect to their computational efficiency during training, by estimating the computational cost of the procedure expressed in floating-point operations. We design a cost model based on analytical expressions and apply it on a collection of 20 KGE models, representative of the state-of-the-art. We show that dimensionality or parameter efficiency, used in the literature to compare models with each other, are not suitable to evaluate the true cost of models. Through fixed-budget experiments, a novel approach to evaluate KGE models based on cost estimates, we re-assess the relative performance of model families compared to the state-of-the-art. Bilinear models such as ComplEx underperform with a low computational budget while hyperbolic linear models appear to offer no particular benefit compared to simpler Euclidian models, especially the MuRE model. Neural models, such as ConvE or CompGCN, achieve reasonable performance in the literature but their high computational cost appears unnecessary when compared with other models. The trade-off between efficiency and expressivity of both linear and neural models is to be further explored.
Cite as

Victor Charpenay, Mansour Zoubeirou A Mayaki, and Antoine Zimmermann. On the Computational Cost of Knowledge Graph Embeddings. In Transactions on Graph Data and Knowledge (TGDK), Volume 4, Issue 1, pp. 1:1-1:30, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@Article{charpenay_et_al:TGDK.4.1.1,
  author =	{Charpenay, Victor and Zoubeirou A Mayaki, Mansour and Zimmermann, Antoine},
  title =	{{On the Computational Cost of Knowledge Graph Embeddings}},
  journal =	{Transactions on Graph Data and Knowledge},
  pages =	{1:1--1:30},
  ISSN =	{2942-7517},
  year =	{2026},
  volume =	{4},
  number =	{1},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/TGDK.4.1.1},
  URN =		{urn:nbn:de:0030-drops-256863},
  doi =		{10.4230/TGDK.4.1.1},
  annote =	{Keywords: Knowledge Graph Embedding, Parameter Efficiency, Computational Budget, Green AI}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.8
Contracts: A Unified Lens on Congestion Control Robustness, Fairness, Congestion, and Generality

Authors: Anup Agarwal, Venkat Arun, and Srinivasan Seshan

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Congestion control algorithms (CCAs) operate in partially observable environments. They cannot directly observe link capacities or competing flows. To share network resources fairly, CCAs (implicitly) communicate fair shares through observable signals. For instance, Reno encodes the fair share as ∝ 1/√{loss rate}. We call such communication mechanisms as contracts. We find that the choice of contract fixes key steady-state performance metrics, including (1) robustness to errors in congestion signals, (2) fairness, (3) amount of congestion (e.g., delay, loss), and (4) generality (e.g., range of supported link rates). This leads to fundamental tradeoffs between these metrics. Further, we show that many contracts lead to starvation (extreme unfairness), and must be avoided. Hence, contracts are a powerful way to analyze tradeoffs and avoid pitfalls in CCA design and analysis. We empirically validate our findings and discuss their implications on CCA design and network measurement.
Cite as

Anup Agarwal, Venkat Arun, and Srinivasan Seshan. Contracts: A Unified Lens on Congestion Control Robustness, Fairness, Congestion, and Generality. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 8:1-8:30, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{agarwal_et_al:OASIcs.NINeS.2026.8,
  author =	{Agarwal, Anup and Arun, Venkat and Seshan, Srinivasan},
  title =	{{Contracts: A Unified Lens on Congestion Control Robustness, Fairness, Congestion, and Generality}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{8:1--8:30},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.8},
  URN =		{urn:nbn:de:0030-drops-255933},
  doi =		{10.4230/OASIcs.NINeS.2026.8},
  annote =	{Keywords: Transport Protocols, Congestion Control, Fairness}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.10
Scalable Routing in a City-Scale Wi-Fi Network for Disaster Recovery

Authors: Ziqian Liu, Om Chabra, James Lynch, Aaron Martin, Chenning Li, and Hari Balakrishnan

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
This paper presents CityMesh, a city-scale decentralized mesh network designed for disaster recovery and emergency scenarios. When wide-area Internet connectivity is unavailable or severely degraded, CityMesh leverages both static access points and mobile devices equipped with Wi-Fi to provide intra-city connectivity and reach opportunistic gateways to the Internet (e.g., via satellite links). The main contribution of this paper is a scalable routing protocol that supports millions of devices, addressing a long-standing limitation of wireless mesh and mobile ad hoc networks. Unlike prior approaches, CityMesh exploits rich building-location and building-geometry data from widely available city maps to guide route computation, improving packet delivery while significantly reducing transmission overhead. Simulation results from 70 cities show that CityMesh improves packet delivery rates by 88% over WEAVE (a state-of-the-art geographic routing protocol). A campus-scale deployment of 300 Wi-Fi devices across 31 buildings shows the practical deployability of CityMesh. These results demonstrate the promise of map-aware routing as a foundation for scalable, resilient city-wide Wi-Fi networks.
Cite as

Ziqian Liu, Om Chabra, James Lynch, Aaron Martin, Chenning Li, and Hari Balakrishnan. Scalable Routing in a City-Scale Wi-Fi Network for Disaster Recovery. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 10:1-10:31, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{liu_et_al:OASIcs.NINeS.2026.10,
  author =	{Liu, Ziqian and Chabra, Om and Lynch, James and Martin, Aaron and Li, Chenning and Balakrishnan, Hari},
  title =	{{Scalable Routing in a City-Scale Wi-Fi Network for Disaster Recovery}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{10:1--10:31},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.10},
  URN =		{urn:nbn:de:0030-drops-255954},
  doi =		{10.4230/OASIcs.NINeS.2026.10},
  annote =	{Keywords: mesh networking, disaster recovery, geographic routing, scalability, Wi-Fi}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.12
No Signal to Rule Them All: A Systematic Analysis of In-Network Congestion Signals

Authors: Sarah McClure, Nandita Dukkipati, Sylvia Ratnasamy, and Scott Shenker

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
In this paper, we address the following question: what in-network signals should a network provide to congestion control algorithms? To answer this guiding question, we use prior work to automatically generate congestion control algorithms optimized for a given performance objective and set of in-network congestion signals. We then make observations about the relative value of these congestion signals across a range of performance objectives. Our analysis yields a surprising central finding: for the average case, sophisticated In-Network Telemetry (INT) offers minimal performance benefits over traditional end-to-end (E2E) signals, with performance typically within 3%. We also find no single "best" INT signal, but rather a clear trade-off that manifests in many scenarios: link-based signals often excel at controlling delay, while queue-based signals are better for maximizing throughput. To make these findings concrete, we validate them by examining the extent to which in-network signals improve the performance of the BBR congestion control algorithm.
Cite as

Sarah McClure, Nandita Dukkipati, Sylvia Ratnasamy, and Scott Shenker. No Signal to Rule Them All: A Systematic Analysis of In-Network Congestion Signals. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 12:1-12:30, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{mcclure_et_al:OASIcs.NINeS.2026.12,
  author =	{McClure, Sarah and Dukkipati, Nandita and Ratnasamy, Sylvia and Shenker, Scott},
  title =	{{No Signal to Rule Them All: A Systematic Analysis of In-Network Congestion Signals}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{12:1--12:30},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.12},
  URN =		{urn:nbn:de:0030-drops-255974},
  doi =		{10.4230/OASIcs.NINeS.2026.12},
  annote =	{Keywords: Congestion control, in-network telemetry}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.15
BISCAY: Practical Radio KPI Driven Congestion Control for Mobile Networks

Authors: Jon Larrea, Tanya Shreedhar, Atte Niemi, Adel Sefiane, and Mahesh K. Marina

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Mobile application performance is often bottlenecked by cellular links with rapid bandwidth fluctuations. We show that radio KPIs from the device chipset can precisely and promptly measure available cellular bandwidth. Building on this, we propose Biscay, a practical KPI-driven congestion control for mobile networks. Biscay leverages OpenDiag, an in-kernel, real-time KPI extractor we introduce along with a KPI-based bandwidth estimator to adjust the congestion window, utilizing available bandwidth while minimizing delay. We implement Biscay and OpenDiag on unrooted Android 5G phones. Across trace-driven emulations and real-world 4G/5G experiments, Biscay outperforms state-of-the-art CCAs (e.g., BBR, CUBIC), typically reducing average and tail delay by >90% while matching or improving throughput. These gains stem from OpenDiag’s 100× finer on-device KPI granularity than existing alternatives like MobileInsight.
Cite as

Jon Larrea, Tanya Shreedhar, Atte Niemi, Adel Sefiane, and Mahesh K. Marina. BISCAY: Practical Radio KPI Driven Congestion Control for Mobile Networks. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 15:1-15:32, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{larrea_et_al:OASIcs.NINeS.2026.15,
  author =	{Larrea, Jon and Shreedhar, Tanya and Niemi, Atte and Sefiane, Adel and Marina, Mahesh K.},
  title =	{{BISCAY: Practical Radio KPI Driven Congestion Control for Mobile Networks}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{15:1--15:32},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.15},
  URN =		{urn:nbn:de:0030-drops-256002},
  doi =		{10.4230/OASIcs.NINeS.2026.15},
  annote =	{Keywords: Cellular Networks, Congestion Control, LTE/5G}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.18
TURBO: Utility-Aware Bandwidth Allocation for Cloud-Augmented Autonomous Control

Authors: Peter Schafhalter, Alexander Krentsel, Hongbo Wei, Joseph E. Gonzalez, Sylvia Ratnasamy, Scott Shenker, and Ion Stoica

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Autonomous driving system progress has been driven by improvements in machine learning (ML) models, whose computational demands now exceed what edge devices alone can provide. The cloud offers abundant compute, but the network has long been treated as an unreliable bottleneck rather than a co-equal part of the autonomous vehicle control loop. We argue that this separation is no longer tenable: safety-critical autonomy requires co-design of control, models, and network resource allocation itself. We introduce TURBO, a cloud-augmented control framework that addresses this challenge, formulating bandwidth allocation and control pipeline configuration across both the car and cloud as a joint optimization problem. TURBO maximizes benefit to the car while guaranteeing safety in the face of highly variable network conditions. We implement TURBO and evaluate it in both simulation and real-world deployment, showing it can improve average accuracy by up to 15.6%pt over existing on-vehicle-only pipelines. Our code is made available at www.github.com/NetSys/turbo.
Cite as

Peter Schafhalter, Alexander Krentsel, Hongbo Wei, Joseph E. Gonzalez, Sylvia Ratnasamy, Scott Shenker, and Ion Stoica. TURBO: Utility-Aware Bandwidth Allocation for Cloud-Augmented Autonomous Control. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 18:1-18:34, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{schafhalter_et_al:OASIcs.NINeS.2026.18,
  author =	{Schafhalter, Peter and Krentsel, Alexander and Wei, Hongbo and Gonzalez, Joseph E. and Ratnasamy, Sylvia and Shenker, Scott and Stoica, Ion},
  title =	{{TURBO: Utility-Aware Bandwidth Allocation for Cloud-Augmented Autonomous Control}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{18:1--18:34},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.18},
  URN =		{urn:nbn:de:0030-drops-256039},
  doi =		{10.4230/OASIcs.NINeS.2026.18},
  annote =	{Keywords: autonomous vehicles, bandwidth allocation, cloud computing, edge computing, machine learning}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.19
Simulate Before Sending: Rethinking Transport in Datacenter Networks

Authors: Dan Straussman, Isaac Keslassy, Alexander Shpiner, and Liran Liss

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Existing transport protocols in commodity datacenter networks struggle to provide low collective completion times (CCTs) to AI training collectives, as packet losses and retransmissions significantly degrade performance. We propose dcSim, an efficient transport that achieves low CCTs and practically lossless performance with commodity switches. In dcSim, each packet first employs a small simulation probe to traverse the network and explore congestion along a candidate path. Only packets whose simulation probes succeed are then transmitted, expecting to succeed as well. Evaluations confirm that dcSim achieves faster CCTs than existing schemes, with small queues and virtually zero packet loss. Finally, dcSim also excels in adverse conditions, including oversubscribed topologies.
Cite as

Dan Straussman, Isaac Keslassy, Alexander Shpiner, and Liran Liss. Simulate Before Sending: Rethinking Transport in Datacenter Networks. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 19:1-19:22, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{straussman_et_al:OASIcs.NINeS.2026.19,
  author =	{Straussman, Dan and Keslassy, Isaac and Shpiner, Alexander and Liss, Liran},
  title =	{{Simulate Before Sending: Rethinking Transport in Datacenter Networks}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{19:1--19:22},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.19},
  URN =		{urn:nbn:de:0030-drops-256044},
  doi =		{10.4230/OASIcs.NINeS.2026.19},
  annote =	{Keywords: Datacenter networks, transport protocols, AI training, lossless networks}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.27
Making Congestion Control Algorithms Insensitive to Underlying Propagation Delays

Authors: Cyrus Illick, Michael Roger, Vishal Misra, and Dan Rubenstein

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Underlying propagation delays, which significantly impact flow Round-Trip-Times (RTTs), fundamentally shape how congestion control algorithms (CCAs) allocate bandwidth among competing flows. Traditionally, lower RTTs were desirable, as a flow rate would decrease as RTT increased under conventional CCAs such as Reno and Cubic. More recently, algorithms like BBR reversed this behavior, favoring flows with larger RTTs. In modern heterogeneous networks, where competing flows may employ different CCAs and serve diverse application workloads, notions of fairness become ambiguous and RTT’s role further complexifies addressing an already very challenging problem. We offer a simple but initially counter-intuitive remedy to address how all CCAs can remove this inherent sensitivity to propagation delay: have all flows "emulate" the same underlying propagation delay. Unsurprisingly, this idea is often met with a lot of resistance, since it is in many ways counter to many fundamental tenets of networking and congestion control. In this paper, we try to make the case that while the idea is not a silver bullet on its own, it can greatly simplify design and predictability across existing and future congestion control protocols going forward.
Cite as

Cyrus Illick, Michael Roger, Vishal Misra, and Dan Rubenstein. Making Congestion Control Algorithms Insensitive to Underlying Propagation Delays. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 27:1-27:26, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{illick_et_al:OASIcs.NINeS.2026.27,
  author =	{Illick, Cyrus and Roger, Michael and Misra, Vishal and Rubenstein, Dan},
  title =	{{Making Congestion Control Algorithms Insensitive to Underlying Propagation Delays}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{27:1--27:26},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.27},
  URN =		{urn:nbn:de:0030-drops-256122},
  doi =		{10.4230/OASIcs.NINeS.2026.27},
  annote =	{Keywords: Congestion Control, Propagation Delay}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.21
Accio: Rethinking OS-Architecture Co-Design for Next-Gen I/O

Authors: Amirmohammad Nazari, Rajit Manohar, and Robert Soulé

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
In this paper, we propose a novel software/hardware design to improve I/O performance while maintaining existing POSIX-standard APIs. Our approach stands in contrast to existing kernel-bypass strategies that improve performance at the expense of abandoning familiar programming abstractions. Our key insight is that navigating the performance-functionality trade-off requires changes to the processor; it cannot be done without support of the CPU micro-architecture. Our design, called Accio, includes: dedicated hardware for interrupt management, a hardware assist for thread scheduling, tables in hardware that manage I/O state, and modifications to the operating system to support the new hardware. Our evaluation demonstrates that Accio saturates the bus bandwidth, reduces CPU usage by up to 66% compared to state-of-the-art kernel-bypass systems, and reduces latency to 1/12th of that of the Linux kernel, matching that of kernel-bypass systems.
Cite as

Amirmohammad Nazari, Rajit Manohar, and Robert Soulé. Accio: Rethinking OS-Architecture Co-Design for Next-Gen I/O. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 21:1-21:24, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{nazari_et_al:OASIcs.NINeS.2026.21,
  author =	{Nazari, Amirmohammad and Manohar, Rajit and Soul\'{e}, Robert},
  title =	{{Accio: Rethinking OS-Architecture Co-Design for Next-Gen I/O}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{21:1--21:24},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.21},
  URN =		{urn:nbn:de:0030-drops-256063},
  doi =		{10.4230/OASIcs.NINeS.2026.21},
  annote =	{Keywords: Networks, Operating Systems, I/O Optimization}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.3
Limix: Limiting Lamport Exposure to Distant Failures in Globally-Managed Distributed Systems

Authors: Cristina Băsescu, Georgia Fragkouli, Enis Ceyhun Alp, Michael F. Nowlan, Jose M. Faleiro, Gaylor Bosson, Kelong Cong, Pierluca Borsò-Tan, Vero Estrada-Galiñanes, and Bryan Ford

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Globalized computing infrastructures offer the convenience and elasticity of globally managed objects and services, but lack the resilience to distant failures that localized infrastructures such as private clouds provide. Providing both global management and resilience to distant failures, however, poses a fundamental problem for configuration services: How to discover a possibly migratory, strongly-consistent service/object in a globalized infrastructure without dependencies on globalized state? Limix is the first metadata configuration service that addresses this problem. With Limix, global strongly-consistent data-plane services and objects are insulated from remote gray failures by ensuring that the definitive, strongly-consistent metadata for any object is always confined to the same region as the object itself. Limix guarantees availability bounds: any user can continue accessing any strongly consistent object that matters to the user located at distance Δ away, insulated from failures outside a small multiple of Δ. We built a Limix metadata service based on the key-value interface of CockroachDB. Our experiments on Internet-like networks and on AWS, using realistic trace-driven workloads, show that Limix enables global management and significantly improves availability over the state-of-the-art.
Cite as

Cristina Băsescu, Georgia Fragkouli, Enis Ceyhun Alp, Michael F. Nowlan, Jose M. Faleiro, Gaylor Bosson, Kelong Cong, Pierluca Borsò-Tan, Vero Estrada-Galiñanes, and Bryan Ford. Limix: Limiting Lamport Exposure to Distant Failures in Globally-Managed Distributed Systems. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 3:1-3:29, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{basescu_et_al:OASIcs.NINeS.2026.3,
  author =	{B\u{a}sescu, Cristina and Fragkouli, Georgia and Alp, Enis Ceyhun and Nowlan, Michael F. and Faleiro, Jose M. and Bosson, Gaylor and Cong, Kelong and Bors\`{o}-Tan, Pierluca and Estrada-Gali\~{n}anes, Vero and Ford, Bryan},
  title =	{{Limix: Limiting Lamport Exposure to Distant Failures in Globally-Managed Distributed Systems}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{3:1--3:29},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.3},
  URN =		{urn:nbn:de:0030-drops-255880},
  doi =		{10.4230/OASIcs.NINeS.2026.3},
  annote =	{Keywords: Distributed systems, Availability, Fault tolerance, Strong-consistency, Coordination systems, Lamport exposure}
}
Document
DOI: 10.4230/OASIcs.NINeS.2026.5
OrbitalBrain: A Distributed Framework for Training ML Models in Space

Authors: Om Chabra, Chenning Li, Kevin Hsieh, Santiago Segarra, Behnaz Arzani, Peder Olsen, and Ranveer Chandra

Published in: OASIcs, Volume 139, 1st New Ideas in Networked Systems (NINeS 2026)

Abstract
Earth observation nanosatellites capture high-resolution photos of the Earth in near real-time. These images increasingly support ML applications that are critical for safety and response, such as forest fire and flood detection. However, the downlink bandwidth is limited, resulting in days or weeks of delay from image capture to training. In this work, we propose OrbitalBrain, an efficient in-space distributed ML training framework that leverages limited and predictable satellite compute, bandwidth, and power to intelligently balance data transfer, model aggregation, and local training. Our evaluations demonstrate that OrbitalBrain achieves 1.52×-12.4× speedup in time-to-accuracy while always reaching a higher final model accuracy compared to state-of-the-art ground-based or federated learning baselines. Furthermore, our approach is complementary to satellite imagery capturing and downloading, enhancing the overall efficiency of satellite-based applications.
Cite as

Om Chabra, Chenning Li, Kevin Hsieh, Santiago Segarra, Behnaz Arzani, Peder Olsen, and Ranveer Chandra. OrbitalBrain: A Distributed Framework for Training ML Models in Space. In 1st New Ideas in Networked Systems (NINeS 2026). Open Access Series in Informatics (OASIcs), Volume 139, pp. 5:1-5:32, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026)

Copy BibTex To Clipboard

@InProceedings{chabra_et_al:OASIcs.NINeS.2026.5,
  author =	{Chabra, Om and Li, Chenning and Hsieh, Kevin and Segarra, Santiago and Arzani, Behnaz and Olsen, Peder and Chandra, Ranveer},
  title =	{{OrbitalBrain: A Distributed Framework for Training ML Models in Space}},
  booktitle =	{1st New Ideas in Networked Systems (NINeS 2026)},
  pages =	{5:1--5:32},
  series =	{Open Access Series in Informatics (OASIcs)},
  ISBN =	{978-3-95977-414-7},
  ISSN =	{2190-6807},
  year =	{2026},
  volume =	{139},
  editor =	{Argyraki, Katerina and Panda, Aurojit},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.NINeS.2026.5},
  URN =		{urn:nbn:de:0030-drops-255907},
  doi =		{10.4230/OASIcs.NINeS.2026.5},
  annote =	{Keywords: Satellite networks, Distributed machine learning, Federated learning, Earth observation, In-orbit computing}
}
  • Refine by Type
  • 42 Document/PDF
  • 36 Document/HTML

  • Refine by Publication Year
  • 15 2026
  • 17 2025
  • 2 2024
  • 6 2023
  • 1 2022
  • Show More...

  • Refine by Author
  • 2 Bonifati, Angela
  • 2 Chabra, Om
  • 2 Chen, Jiaoyan
  • 2 Curzel, Serena
  • 2 Hose, Katja
  • Show More...

  • Refine by Series/Journal
  • 16 LIPIcs
  • 14 OASIcs
  • 1 LITES
  • 11 TGDK

  • Refine by Classification
  • 5 Information systems → Graph-based database models
  • 5 Networks → Transport protocols
  • 3 Computing methodologies → Knowledge representation and reasoning
  • 2 Computer systems organization → Distributed architectures
  • 2 Computer systems organization → Reconfigurable computing
  • Show More...

  • Refine by Keyword
  • 3 Congestion Control
  • 3 Knowledge graphs
  • 3 Large Language Models
  • 2 Knowledge Graphs
  • 2 LLM
  • Show More...

Any Issues?
X

Feedback on the Current Page

CAPTCHA

Thanks for your feedback!

Feedback submitted to Dagstuhl Publishing

Could not send message

Please try again later or send an E-mail
© 2023-2026 Schloss Dagstuhl – LZI GmbH Schloss Dagstuhl – LZI GmbH About DROPS Imprint Privacy Contact