AFROSCOPE identifies 640 African languages, beats frontier LLMs' 47.4 F1 with 97.9
mageed · x · 2026-10-27
A UBC team presented AFROSCOPE at EMNLP 2026, a framework for African language identification covering 640 languages across 8 families and 7 scripts, trained on 5.95M deduplicated sentences. Bible-derived text was cut from 73-95% to 53.5%, and the corpus has zero overlap with all six evaluation benchmarks (a widely used resource overlaps up to 14.15%).
Results: SerengetiAF + Mirror hits 97.9 macro-F1 on the AFROSCOPE test set vs 71.9 for GlotLID-M; the best frontier LLM (Claude Opus 4.7) manages only 47.4. A hierarchical second-pass 'Mirror' that separates confusable languages adds +1.57, and the authors find script—not language family—best predicts transfer.
More from Research
- S3AC: Shanghai AI Lab's Stabilized Actor-Critic Method for LLM Long-Chain Reasoning — Shanghai-AI-Laboratory · 2026-10-27
- Skill∞: Recursive Self-Improvement Where Agents Optimize Their Own Optimization Skills — NationalUniversityofSingapore · 2026-10-27
- 229k-sample replication of Cohere's Qwen 3.8 malicious PUT claim raises more questions — builderjaydub · 2026-10-27
- Physicist on OpenAI's math result: it's mathematical physics, not theoretical physics — skdh · 2026-10-27
- Mechanistic Interpretability and Fast Memory Recall Are the Same Problem — mayfer · 2026-10-27
- Tsinghua team open-sources C2C: LLMs swap KV-Caches directly for 2.5x faster, more accurate collaboration — solyarisoftware · 2026-10-27