Research map / Label-efficient and robust learning / Recognition and long-tail
Filtered Inner Product Projection for Crosslingual Embedding Alignment
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add self-learning with 14K additional pairs (ablation) | FIPP · XLING 1K BLI, all 28 language pairs | MAP: 0.344 -> 0.406 | Helped |
| Add self-learning with 14K additional pairs (ablation) | FIPP · XLING 5K BLI, all 28 language pairs | MAP: 0.442 -> 0.441 | No clear effect |
| Add isotropic preprocessing (ablation) | Procrustes-based FIPP ablation · XLING 1K BLI, all 28 language pairs | MAP: 0.333 -> 0.344 | Helped |
| Add isotropic preprocessing (ablation) | Procrustes-based FIPP ablation · XLING 5K BLI, all 28 language pairs | MAP: 0.430 -> 0.442 | Helped |
| Add inner product filtering (ablation) | Procrustes-based FIPP ablation · XLING 1K BLI, all 28 language pairs | MAP: 0.335 -> 0.344 | Helped |
| Add inner product filtering (ablation) | Procrustes-based FIPP ablation · XLING 5K BLI, all 28 language pairs | MAP: 0.439 -> 0.442 | Helped |
| Add weighted Procrustes (ablation) | Procrustes-based FIPP ablation · XLING 1K BLI, all 28 language pairs | MAP: 0.336 -> 0.344 | Helped |
| Add weighted Procrustes (ablation) | Procrustes-based FIPP ablation · XLING 5K BLI, all 28 language pairs | MAP: 0.440 -> 0.442 | Helped |
| Run multiple FIPP optimization iterations instead of one (ablation) | FIPP · XLING 1K and 5K BLI | MAP: 0.002; 0.001; 0.002; 0.002 | No clear effect |
| Use FIPP rather than the strongest reported baseline | Bilingual lexicon induction · XLING 1K BLI, all 28 language pairs | MAP: 0.379 -> 0.406 | Helped |
| Use FIPP rather than the strongest reported baseline | Bilingual lexicon induction · XLING 5K BLI, all 28 language pairs | MAP: 0.437 -> 0.442 | Helped |
| Use FIPP instead of RCSLS | Cross-lingual document classification · TED-CLDC, average | micro-averaged F1: 0.510 -> 0.477 | Hurt |
Closest papers by meaning
- Enhancing Bilingual Lexicon Induction via Bi-directional Translation Pair Retrieving (AAAI 2024 · similarity 0.87)
- Constrained Decoding for Cross-lingual Label Projection (ICLR 2024 · similarity 0.85)
- MMTEB: Massive Multilingual Text Embedding Benchmark (ICLR 2025 · similarity 0.85)
- DEPT: Decoupled Embeddings for Pre-training Language Models (ICLR 2025 · similarity 0.85)
- The Scandinavian Embedding Benchmarks: Comprehensive Assessment of Multilingual and Monolingual Text Embedding (NeurIPS 2024 · similarity 0.84)
- CL2CM: Improving Cross-Lingual Cross-Modal Retrieval via Cross-Lingual Knowledge Transfer (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.