GLiClass multilang ultra

Name: GLiClass multilang ultra
Author: Knowledgator

KnowledgatorOpen weights

The multilingual GLiClass checkpoint.

Specifications

Configuration	Mode	Decisions	Beats baseline	ECE
criteria + single-hypothesis Noul (correct payload)headline	serial	0.628	4/7	0.185
criteria folded into labels	serial	0.422	3/7	0.434
default payload	serial	0.402	3/7	0.513

Per-question accuracy — criteria + single-hypothesis Noul (correct payload), serial mode, 120 judgments

Task	Question	Type	Accuracy	Baseline	Beats?
agent_traces	outcome	choice	0.500	0.357	✅
agent_traces	needs_human	noul	0.714	0.714	❌
customer_service	department	choice	0.750	0.400	✅
customer_service	refund_requested	noul	0.950	0.700	✅
customer_service	frustration	score	0.350	0.500	❌
security_incidents	severity	choice	0.500	0.312	✅
security_incidents	actively_exploited	noul	0.688	0.688	❌

Per-question accuracy — correct payload, no prompt, serial mode, 2,000 judgments

Task	Question	Type	Accuracy	Baseline	Beats?
td_agent_trace_observability	action	choice	0.310	0.360	❌
td_agent_trace_observability	needs_review	noul	0.510	0.550	❌
td_agent_trace_observability	outcome	choice	0.210	0.500	❌
td_agent_trace_observability	risk	score	0.340	0.340	❌
td_agent_trace_observability	urgency	score	0.330	0.370	❌
td_customer_service	action	choice	0.320	0.520	❌
td_customer_service	category	choice	0.880	0.380	✅
td_customer_service	churn_risk	score	0.650	0.470	✅
td_customer_service	needs_human	noul	0.330	0.670	❌
td_customer_service	urgency	score	0.580	0.420	✅
td_invoice_processing	discrepancy_severity	score	0.070	0.470	❌
td_invoice_processing	disposition	choice	0.450	0.450	❌
td_invoice_processing	duplicate	noul	0.110	0.890	❌
td_invoice_processing	matches_order	noul	0.540	0.540	❌
td_invoice_processing	urgency	score	0.320	0.510	❌
td_security_incidents	credential_compromise	noul	0.440	0.550	❌
td_security_incidents	disposition	choice	0.080	0.740	❌
td_security_incidents	severity	score	0.380	0.490	❌
td_security_incidents	true_positive	noul	0.670	0.730	❌
td_security_incidents	urgency	score	0.380	0.500	❌

Fitted temperature per primitive (fold 0, judged on fold 1): choice 5, score 4, noul 7 — ECE 0.273 → 0.031, accuracy → 0.370.

How does GLiClass multilang ultra compare to Jev?

On typed-decisions GLiClass multilang ultra scores 0.395 against Jev's 0.726, beating the majority baseline on 3 of 20 questions (Jev: 18/20).

Is GLiClass multilang ultra open source?

Yes. knowledgator/gliclass-multilang-ultra is released under Apache 2.0 and runs on your own hardware with the gliclass library.

Can GLiClass multilang ultra extract entities?

No. GLiClass multilang ultra answers questions about text and never returns spans; use a GLiNER-family model for extraction.