Research map / Learning theory and algorithms / Fairness
Experiential Fairness: Bridging the Gap Between User Experience and Resource-Centric Fairness in Online LLM Services
Jiahua Huang, Wentai Wu, Yongheng Liu, Guozhi Liu, Yang Wang, Weiwei Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| schedule by SLO-aware fairness index with credit exchange (ExFairS) instead of FCFS, RR or VTC | multi-tenant LLM serving · six workload scenarios | SLO violation rate: reduces by up to 100% | Improved |
| schedule with ExFairS instead of baseline schedulers | multi-tenant LLM serving · six workload scenarios | system throughput (tokens/s): +14-21.9% | Improved |
| use SAFI-based credit exchange instead of max-min QoE adapted scheduler | ExFairS · Scenario V (20 clients) | converged SAFI: lowest SAFI about 0.17 | Improved |
| lower weighting factor alpha in SAFI (less emphasis on user experience) (ablation) | ExFairS · Scenario III | SLO compliance | Mixed |
| Round Robin versus ExFairS under heterogeneous SLOs | Round Robin · Scenario IV | normalized latency score: RR score 0.00 | Hurt |
| FCFS under 50 clients | FCFS · Scenario VI | Jain's Index vs core metrics | Mixed |
Closest papers by meaning
- Evaluating Test-Time Scaling of General LLM Agents (NeurIPS 2026 · similarity 0.83)
- SpecEdge: Scalable Edge-Assisted Serving Framework for Interactive LLMs (NeurIPS 2025 · similarity 0.83)
- Justitia: Fair and Efficient Scheduling of Task-parallel LLM Agents with Selective Pampering (NeurIPS 2026 · similarity 0.83)
- Beyond Prediction: Tail-Aware Scheduling for LLM Inference (ICML 2026 · similarity 0.83)
- MAPS: Memory-Aware Predictive Scheduling Framework for Large Language Model Serving (ICML 2026 · similarity 0.82)
- OServe: Accelerating LLM Serving via Spatial-Temporal Workload Orchestration (ICML 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
