Spotlight

Papers

Field

Top tools here

Open problems in Federated learning

Heterogeneous preferences and gradient conflicts in federated personalization1 paper raises it · 1 addresses it
Diverse user objectives and scarce feedback make it hard to learn shared initializations that adapt well to individual preferences.
- Collaborative Personalized Preference Alignment for LLMs under Data Deficiencyraises · addresses
Decentralized inference-time collaboration without global coordination1 paper raises it · 1 addresses it
Exploiting distributed expertise in decentralized federated learning requires efficient, local-only routing and aggregation.
- Learning to Refer: Inference-Time Collaboration in Decentralized Federated Learningraises · addresses
Communication-efficient federated learning under non-IID data and client drift1 paper raises it · 1 addresses it
Bandwidth constraints and non-IID data cause high latency and client drift, degrading model performance.
- Communication-Efficient Federated Domain Generalization in Edge Networks via Dynamic Prototype Regulationraises · addresses
Secure aggregation with adaptive user dropouts1 paper raises it · 1 addresses it
Existing secure aggregation schemes cannot dynamically adjust to actual dropout scenarios, limiting practicality.
- Fundamental Limits of Information-Theoretic Secure Aggregation with User Dropouts in Federated Learningraises · addresses
Federated continual learning with heterogeneous device capacities1 paper raises it · 1 addresses it
Training models of varying sizes on devices with drifting data streams is underexplored and challenging.
- DROD: Dual-Ratio Ordered Dropout for Federated Continual Learning on Heterogeneous Devicesraises · addresses

Named by an AI from these papers' abstracts.

Open problems

Recently funded

Topics in this field

differential privacy · private learning · differentially private90 papers21 this week
federated learning · federated learning fl · distributed learning97 papers19 this week

Topics in this field

Active contributors · 90 days

Mohamed ElmahallawyVeil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction4
Mohamed ShaabanVeil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction4
Hassan AshtianiPrivate Evaluation of High-Dimensional Functions2
Mohammad AfzaliPrivate Evaluation of High-Dimensional Functions2
Christos AnagnostopoulosLocal Epochs as Optimal Stopping: Theory and Evidence in Federated Learning2
Ke XiaoLocal Epochs as Optimal Stopping: Theory and Evidence in Federated Learning2

Active contributors

Activity · 12 weeks

Coming with more sources: repos

Activity

★ Spotlight this month

arXiv cs.AIOct 15

Private Retrieval Augmented Generation via Random Projection

Dixi Yao, Tahseen Rabbani, Tian Li

Retrieval-Augmented Generation (RAG) enhances the capabilities of large language models (LLMs) by querying external structured knowledge. However, it can also introduce privacy risks by leaking sensitive information from the retrieval datastore. We propose a simple method to preserve datastore privacy in RAG systems via random projection. By applying the same projection to both datastore embeddings and query embeddings, our method provably preserves semantic similarity between queries and retrieved items while substantially mitigating data extraction attacks. Across multiple RAG architectures and datasets, we show that this lightweight approach achieves superior retrieval and generation performance compared to prior methods which leverage formal differential privacy (DP) guarantees, while exhibiting comparable empirical privacy under strong attack models. Our results for the first time suggest that random projection can serve as a competitive and practical baseline for privacy-preserving RAG systems.

PDF arXiv page Details

Spotlight

Recent papers

All in Research →

No papers with code in this field yet.

Papers

Topics in this field

differential privacy · private learning · differentially private90 papers21 this week
federated learning · federated learning fl · distributed learning97 papers19 this week

★ Spotlight this month

arXiv cs.AIOct 15

Private Retrieval Augmented Generation via Random Projection

Dixi Yao, Tahseen Rabbani, Tian Li

Retrieval-Augmented Generation (RAG) enhances the capabilities of large language models (LLMs) by querying external structured knowledge. However, it can also introduce privacy risks by leaking sensitive information from the retrieval datastore. We propose a simple method to preserve datastore privacy in RAG systems via random projection. By applying the same projection to both datastore embeddings and query embeddings, our method provably preserves semantic similarity between queries and retrieved items while substantially mitigating data extraction attacks. Across multiple RAG architectures and datasets, we show that this lightweight approach achieves superior retrieval and generation performance compared to prior methods which leverage formal differential privacy (DP) guarantees, while exhibiting comparable empirical privacy under strong attack models. Our results for the first time suggest that random projection can serve as a competitive and practical baseline for privacy-preserving RAG systems.

PDF arXiv page Details

Recent papers

All in Research →

No papers with code in this field yet.

Active contributors · 90 days

Mohamed ElmahallawyVeil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction4
Mohamed ShaabanVeil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction4
Hassan AshtianiPrivate Evaluation of High-Dimensional Functions2
Mohammad AfzaliPrivate Evaluation of High-Dimensional Functions2
Christos AnagnostopoulosLocal Epochs as Optimal Stopping: Theory and Evidence in Federated Learning2
Ke XiaoLocal Epochs as Optimal Stopping: Theory and Evidence in Federated Learning2

Activity · 12 weeks

Coming with more sources: repos

Fields · AI · updated 36m ago

Federated learning

Found by the topic model: federated learning, differential privacy.

40papers this week+8.1% vs last week

All papers · 187 With code · 0

Newest Most relevant

179 papers this month

Open problems in Federated learning

Heterogeneous preferences and gradient conflicts in federated personalization1 paper raises it · 1 addresses it
Diverse user objectives and scarce feedback make it hard to learn shared initializations that adapt well to individual preferences.
- Collaborative Personalized Preference Alignment for LLMs under Data Deficiencyraises · addresses
Decentralized inference-time collaboration without global coordination1 paper raises it · 1 addresses it
Exploiting distributed expertise in decentralized federated learning requires efficient, local-only routing and aggregation.
- Learning to Refer: Inference-Time Collaboration in Decentralized Federated Learningraises · addresses
Communication-efficient federated learning under non-IID data and client drift1 paper raises it · 1 addresses it
Bandwidth constraints and non-IID data cause high latency and client drift, degrading model performance.
- Communication-Efficient Federated Domain Generalization in Edge Networks via Dynamic Prototype Regulationraises · addresses
Secure aggregation with adaptive user dropouts1 paper raises it · 1 addresses it
Existing secure aggregation schemes cannot dynamically adjust to actual dropout scenarios, limiting practicality.
- Fundamental Limits of Information-Theoretic Secure Aggregation with User Dropouts in Federated Learningraises · addresses
Federated continual learning with heterogeneous device capacities1 paper raises it · 1 addresses it
Training models of varying sizes on devices with drifting data streams is underexplored and challenging.
- DROD: Dual-Ratio Ordered Dropout for Federated Continual Learning on Heterogeneous Devicesraises · addresses

Named by an AI from these papers' abstracts.