Jev Decision Models Hit 99.8% of RIC 1s Budget in 6G Open RAN While Hosted LLMs Fall to 0%
Delong Li · hf · 2026-10-06
- Setting: Intent-based Open RAN needs an interpreter turning intents into A1 policies within the RIC loop. Decision models like Jev-1.13.0 return typed policy fields; LLMs generate policy token by token.
- Setup: Compared Jev-1.13.0 and two other decision models against LLMs on the RANIntent v1 benchmark, in closed-loop ns-3 simulation, and on a real A1/E2 path.
- Latency: LLM interpretation takes a median 0.286–2.35 s vs under 25 ms for A1/E2 transfer; Jev-1.13.0 meets the 1 s near-real-time budget on 99.8% of calls, while two hosted LLMs manage 17.9% and 0%.
- Radio performance: Ideal enforcement moves affected-class SLA violations by 3.96 percentage points in the intent's direction; no hosted LLM showed a resolved improvement over Jev-1.13.0, and per-second direct control gave no resolved gain over a numerical xApp.
- Conclusion: Slow interpreters miss the 1 s budget and two saturate queues at 2 intents/s, though no radio penalty of slow interpreters was resolved at the base point.
More from Models
- OpenAI models bypassed isolation controls; governance expert parses recent AI safety incidents — LuizaJarovsky · 2026-10-06
- Debate Continues: Astra Matches Opus 5.5 With Roughly 5x Fewer Reasoning Tokens — VraserX · 2026-10-06
- Microsoft briefly confirms OpenAI uses Looped Transformers in GPT-6 series, then scrubs the page — ResearchCrafty1804 · 2026-10-06
- X's open-source recommendation algorithm adds a NOTICE visibility outcome and new watch-time signals — tetsuoai · 2026-10-06
- Mistral to unveil new flagship model that claims to beat Chinese models on cyber, says Reuters — testingcatalog · 2026-10-06
- Local Qwen3.8-Flash-Next vs Claude Opus 5.5 on the same Rust feature: 7x slower, free, better tests — deepu105 · 2026-10-06