PERIT-S2S · Telecom

Plan changes, SIM swaps and outage triage, mostly over mobile lines that drop mid-sentence.

Audio comes from inbound support calls, retention desks, field escalations. Scored on task success, with a working telecom operator as the control.

1,240
Clips in the target held-out set
96h
Audio, never published
7
Languages in the set
1
Working telecom operator as control

Leaderboard

Illustrative figures. Public runs begin when the held-out sets close.
ModelTask success (%)Median turn latencyBarge-in recovery
Human control · sector agentHuman control
93.3 ± 2.0
623ms95.3%
Frontier realtime · lab AFrontier
54.4 ± 1.7
697ms64.6%
Frontier realtime · lab BFrontier
54.0 ± 1.2
1031ms68.2%
Cascaded stack (ASR → LLM → TTS)Specialist
48.4 ± 1.8
1213ms49.1%
Frontier multimodal · lab CFrontier
44.4 ± 0.9
1027ms48.9%
Open weights · 8B speechOpen weights
29.7 ± 0.8
1352ms32.5%
Task families

What the set actually asks for.

  • Turn-taking & barge-in
  • Task completion
  • Instruction adherence
  • Disambiguation & repair
  • Escalation & refusal
  • Tone under pressure
Method & limits

How a run on this set is produced.

  1. 01Callers are scripted personas played by trained speakers, not synthesised — including the ones who talk over the agent.
  2. 02Every call is scored against the rubric by two reviewers; a third adjudicates splits.
  3. 038 calls per persona per model. Reported intervals are bootstrapped at 95%.
  4. 04Human control is a licensed agent from that sector handling the same personas blind.

Limits, stated up front: 7 languages only, single-channel audio, and no cross-sector transfer. The control is one operator per sector, so the human line carries its own error bar — it is a working standard, not a ceiling.