Communication Efficiency
Cross-source consensus on Communication Efficiency from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Benefits
Comparisons
Highlighted claims
- Fed-FSTQ reached a 60 percent target validation accuracy with 46 times less cumulative uplink traffic than standard Fed-LoRA. — FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
- Fed-FSTQ reduced heterogeneous-uplink straggler-limited time-to-accuracy by 52 percent. — FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
- Fed-FSTQ added 0.85 seconds of Fisher overhead per round but reduced communication by 354.40 seconds. — FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
- The latency and energy gains are attributed to a regime where communication time dominates computation time. — FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices
- The energy advantage of Fed-FSTQ persisted across different transmit-power assumptions. — FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices