AI Paper Atlas

Research map / Large language models / Agents, code and math

NetArena: Dynamically Generated LLM Benchmarks for Network Applications

Yajie Zhou, Jiajun Ruan, Eric S. Wang, Sadjad Fouladi, Francis Y. Yan, Kevin Hsieh, Zaoxing Liu

ICLR 2026 · Large language models › Agents, code and math · approach: Software engineering agents

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 changedWhereMeasuredEffect
scale query set from 100-150 to 2000-5000 queriesNetArena evaluation · CP, Routing, K8sconfidence-interval overlap between agents: 85% -> 0%Improved
use larger query set (5000 vs 100 queries)GPT+ReAct vs Qwen agents · Capacity planningCI overlap: more than 50% overlap -> noneImproved
use K8s:2000 instead of K8s:150 queriesGPT+ReAct · K8ssafety rateHurt
increase task complexity from Level 1 to Level 3GPT-4o+Few-shot · Capacity planningcorrectness and safetyHurt
fine-tune on single complexity level only (SFT-Level-2) (ablation)Qwen-7B · Capacity planning Level 3correctnessHurt
fine-tune on mixed-level data (2000 queries) instead of one level (ablation)Qwen-7B · Capacity planningcorrectness: over 0.96 across all levelsImproved
single-level SFT effect on safety across levels (ablation)Qwen-7B · Capacity planningsafety rateNo clear effect
GRPO fine-tuning with emulator rewardsQwen2.5-0.5B · Routing (Mininet)valid command generationImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy