3arXiv cs.CL (Computation and Language)·
CAMMAR: Culture-Aware Matryoshka embeddings for Arabic metaphor detection
Researchers introduce CAMMAR, a representation learning framework that organizes Arabic language meaning into nested lexical, cultural, and metaphorical embedding subspaces using a staged semantic curriculum inspired by Al-Jurjani's classical theory of nazum. The framework addresses 'semantic smearing' in current Arabic LMs and yields a training-free geometric metaphoricity measure. On a new span-annotated Arabic metaphor dataset, the geometric readout achieves AUC up to 0.84 and correctly ranks figurative over literal counterparts in 82.6% of pairs under paired supervision, though performance collapses to chance under unsupervised domain contrast alone.
Related guides (1)
Related events (8)
Introduction to Matryoshka Embedding Models
This Hugging Face blog post introduces Matryoshka Representation Learning (MRL), a technique for training embedding models that encode information at multiple granularities within a single vector. The approach allows truncating embeddings to smaller dimensions without significant loss in retrieval quality, enabling flexible trade-offs between storage/compute costs and accuracy. The post covers training, evaluation, and practical usage of Matryoshka embedding models via the Sentence Transformers library.
Symbal: Automated detection of systematic misalignments in MLLM-generated image captions
Researchers from Stanford introduce Symbal, a dual-stage framework using off-the-shelf foundation models to detect recurring, visually-correlated errors in multimodal LLM-generated captions. They also release SymbalBench, a benchmark of 1.7 million image-text pairs across 420 vision-language datasets covering natural and medical imaging domains. Symbal correctly identifies systematic misalignments in 63.8% of datasets, nearly 4x better than the closest baseline, and works without access to the underlying MLLM. The work addresses a practical data-quality auditing problem relevant to anyone building or evaluating vision-language datasets.
LangMAP: Language-adaptive tokenization from a shared vocabulary without language identification at inference
LangMAP (Language-adaptive Maximum a Posteriori Tokenization) extends the UnigramLM algorithm to produce language-specific tokenizations from a single shared vocabulary, eliminating the need to retrain models or swap vocabularies for multilingual settings. A key property is that language labels are only required at training time; inference proceeds without language identification. Evaluated across 14 tokenizers, 9 natural languages, and 9 programming languages, LangMAP improves morphological boundary alignment and AST-leaf alignment for all coding languages tested. Fine-tuning results are mixed: consistent gains on grammatical acceptability (MultiBLiMP) but less consistent on knowledge tasks (Global-PIQA, Belebele).
Transformer embeddings shown to intrinsically encode Russell's circumplex model of emotion geometry
A new arXiv paper investigates whether Transformer-based text and speech encoders (RoBERTa, wav2vec 2.0) recover the geometric structure of Russell's circumplex model of affect — a valence-arousal topology from psychology. Experiments on naturalistic datasets (MSP-Podcast) and LLM-generated stimuli show that multimodal fusion achieves perfect topological alignment with Russell's primary emotion ordering, and zero-shot generic text embeddings place fine-grained emotion terms near their human-mapped coordinates. The authors argue this structure is intrinsically encoded in the representations rather than being an artifact of labeling, bridging psychological theory and representation learning.
Meta releases Llama 4 Maverick 17B-128E multimodal instruct model on Hugging Face
Meta released Llama 4 Maverick, a 17B active parameter model with 128 experts (MoE architecture), as an image-text-to-text instruct model on Hugging Face. The model supports multimodal inputs and multiple languages including Arabic, German, and English. With 28K+ downloads and 493 likes shortly after release, it is seeing significant early adoption.
Alyah: Benchmark for Evaluating Emirati Dialect Capabilities in Arabic LLMs
TII UAE introduces Alyah, a benchmark designed to evaluate large language models on Emirati Arabic dialect understanding and generation. The work addresses a gap in Arabic NLP evaluation, where most benchmarks focus on Modern Standard Arabic and neglect regional dialects. The benchmark aims to provide robust assessment of LLM capabilities specific to Emirati linguistic and cultural context.
MATCHA: Contrastive Semantic Alignment Metric for LLM Evaluation
MATCHA is a new automatic evaluation metric for LLMs that addresses a fundamental flaw in existing metrics: both token-overlap (ROUGE) and embedding-based (BERTScore) metrics routinely assign near-identical scores to semantically contradictory texts. The metric uses a dual-view approach that rewards proximity to a gold reference while penalizing adversarially generated counterfactual contradictions. Evaluated across eight benchmarks spanning QA, summarization, NLI, and semantic similarity tasks, MATCHA outperforms 23 embedding models and achieves 18.38% and 20.82% improvements over ROUGE-L and BERTScore respectively on TruthfulQA. Code and metric are publicly released.
AnyMo: Geometry-Aware Setup-Agnostic Framework for Wearable IMU Human Motion Understanding
AnyMo is a geometry-aware framework that addresses the setup-dependence problem in wearable IMU-based human motion modeling by using physics-grounded simulation over dense body-surface placements to generate synthetic training signals. It pre-trains a graph encoder from synthetic placement views and masked partial observations, then tokenizes multi-position IMU data into full-body motion tokens aligned with an LLM for motion-language understanding. Evaluated across zero-shot activity recognition (14 unseen datasets), cross-modal retrieval, and motion captioning, AnyMo improves average Accuracy/F1 by ~11.7%/11.6%, zero-shot retrieval MRR by 15.9–28.6%, and captioning BERT-F1 by 18.8%. The work positions itself as a generalist model for wearable motion understanding transferable across devices and sensing configurations.