Label uncertain forecasts
Help the model learn
The Voidly Sentinel classifier's lowest-confidence forecasts. The model is most uncertain about these 45 predictions — human labels here have the highest information value for the next retrain. Submit observed outcomes via the auth-gated /v1/sentinel/report_miss endpoint.
Method: Impact-aware ranking — Settles 2009 (UW TR-1648). Uncertainty fallback — Lewis 1994. Calibrated uncertainty — arXiv 2510.03162 (2025). · Updated every 15 min · Raw JSON
How to submit a label
How to submit a label
- Get an API key — admin via Voidly, or subscriber key via /alerts.
- POST to
/v1/sentinel/report_misswith:
curl -X POST https://api.voidly.ai/v1/sentinel/report_miss \
-H "X-Voidly-Subscriber-Key: $YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"country_code": "IR",
"what_happened": "Observed shutdown 2026-05-19 14:00 UTC, BGP withdrawal of AS31549. Sourced from NetBlocks twitter + my SSH probe in Tehran.",
"source_url": "https://twitter.com/netblocks/status/..."
}'Rate limit: 10 / hour / key. The label enters error_queue on the Vultr ML server and is incorporated by the next nightly retrain. See /methodology for the gate logic that decides whether new models replace the serving one.
50 uncertain forecasts (lowest distance from 0.5 first)
- FRFRDate unavailablenot yet resolvedProbability50.2%Distance from 0.50.2 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- RussiaRUDate unavailablenot yet resolvedProbability49.4%Distance from 0.50.6 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- IranIRDate unavailablenot yet resolvedProbability50.4%Distance from 0.50.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- FRFRDate unavailablenot yet resolvedProbability50.7%Distance from 0.50.7 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- FRFRDate unavailablenot yet resolvedProbability50.9%Distance from 0.50.9 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- RussiaRUDate unavailablenot yet resolvedProbability48.6%Distance from 0.51.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- RussiaRUNov 18, 2026not yet resolvedProbability48.3%Distance from 0.51.7 ppPred above 0.5?— novery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- RussiaRUDate unavailablenot yet resolvedProbability51.7%Distance from 0.51.7 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- FRFRDate unavailablenot yet resolvedProbability48.6%Distance from 0.51.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- IranIROct 29, 2026resolved wrongProbability51.6%Distance from 0.51.6 ppPred above 0.5?— novery_close_to_threshold · already_resolved_as_wrong · high_volume_country · recent_drift
- FRFRDate unavailablenot yet resolvedProbability51.7%Distance from 0.51.7 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- SESEDate unavailablenot yet resolvedProbability50.0%Distance from 0.50.0 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- SESEDate unavailablenot yet resolvedProbability48.8%Distance from 0.51.2 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- SESEDate unavailablenot yet resolvedProbability48.8%Distance from 0.51.2 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- MyanmarMMNov 12, 2026not yet resolvedProbability48.8%Distance from 0.51.2 ppPred above 0.5?— novery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- IndiaINDate unavailablenot yet resolvedProbability50.6%Distance from 0.50.6 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- IndiaINNov 15, 2026not yet resolvedProbability48.3%Distance from 0.51.7 ppPred above 0.5?— novery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- PakistanPKDate unavailablenot yet resolvedProbability50.3%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- CubaCUNov 17, 2026not yet resolvedProbability50.0%Distance from 0.50.0 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · recent_drift
- HNHNDate unavailablenot yet resolvedProbability50.3%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · recent_drift
- KHKHDate unavailablenot yet resolvedProbability49.1%Distance from 0.50.9 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country · recent_drift
- OMOMDate unavailablenot yet resolvedProbability50.3%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- VenezuelaVEOct 29, 2026resolved wrongProbability49.2%Distance from 0.50.8 ppPred above 0.5?— novery_close_to_threshold · already_resolved_as_wrong · high_volume_country
- EgyptEGOct 28, 2026resolved wrongProbability50.4%Distance from 0.50.4 ppPred above 0.5?— novery_close_to_threshold · already_resolved_as_wrong · high_volume_country
- IDIDNov 1, 2026resolved correctProbability50.4%Distance from 0.50.4 ppPred above 0.5?↑ yesvery_close_to_threshold · already_resolved_as_correct · high_volume_country
- TurkeyTRDate unavailablenot yet resolvedProbability50.4%Distance from 0.50.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- TurkeyTRNov 13, 2026not yet resolvedProbability49.3%Distance from 0.50.7 ppPred above 0.5?— novery_close_to_threshold · not_yet_resolved · high_volume_country
- TurkeyTRNov 1, 2026resolved wrongProbability48.4%Distance from 0.51.6 ppPred above 0.5?— novery_close_to_threshold · already_resolved_as_wrong · high_volume_country
- EthiopiaETDate unavailablenot yet resolvedProbability50.3%Distance from 0.50.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- DODODate unavailablenot yet resolvedProbability50.3%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- DODODate unavailablenot yet resolvedProbability48.3%Distance from 0.51.7 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- UzbekistanUZDate unavailablenot yet resolvedProbability51.2%Distance from 0.51.2 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- GHGHDate unavailablenot yet resolvedProbability48.8%Distance from 0.51.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- CMCMDate unavailablenot yet resolvedProbability51.0%Distance from 0.51.0 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- CMCMDate unavailablenot yet resolvedProbability51.0%Distance from 0.51.0 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- CMCMDate unavailablenot yet resolvedProbability51.1%Distance from 0.51.1 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- CMCMDate unavailablenot yet resolvedProbability51.6%Distance from 0.51.6 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- KEKEDate unavailablenot yet resolvedProbability49.8%Distance from 0.50.2 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- LYLYDate unavailablenot yet resolvedProbability49.7%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- GEGEDate unavailablenot yet resolvedProbability50.5%Distance from 0.50.5 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- GEGEDate unavailablenot yet resolvedProbability51.0%Distance from 0.51.0 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- MAMADate unavailablenot yet resolvedProbability49.5%Distance from 0.50.5 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- CICIDate unavailablenot yet resolvedProbability49.7%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- CICIDate unavailablenot yet resolvedProbability49.7%Distance from 0.50.3 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- ITITDate unavailablenot yet resolvedProbability48.3%Distance from 0.51.7 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
- CICIDate unavailablenot yet resolvedProbability50.5%Distance from 0.50.5 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- HTHTDate unavailablenot yet resolvedProbability50.6%Distance from 0.50.6 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- DZDZDate unavailablenot yet resolvedProbability48.6%Distance from 0.51.4 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- HRHRDate unavailablenot yet resolvedProbability50.5%Distance from 0.50.5 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved
- NONODate unavailablenot yet resolvedProbability50.8%Distance from 0.50.8 ppPred above 0.5?↑ yesvery_close_to_threshold · not_yet_resolved · high_volume_country
Why active learning matters
Most ML systems retrain on random labeled samples. Active learning literature shows that labeling the lowest-confidence predictions instead gives the biggest F1 gain per labeling-budget unit.
Voidly's weekly classifier retrain has been rejecting new model candidates since 2026-05-03 because they F1-regress on a dedup'd evaluation set. The gate is working as designed, but without fresh labels we can't distinguish “real regression” from “data drift the model needs to learn.”
This page closes that loop. Journalists + censorship experts see the model's most uncertain forecasts, submit observed outcomes, and the next retrain gate uses their labels for evaluation. Slow but durable.
References
References
- Calibrated Uncertainty Sampling for Active Learning (arXiv 2510.03162, 2025)
- RiskID: Active learning for network traffic
- /sentinel/backtest — honest model reliability
- /atlas/models — full model registry
- /methodology — retrain gate logic