Multimodal fake news proliferates on social media, yet its verification is hindered by context sparsity and analytical heterogeneity: deceptive posts often hinge on implicit social background, while distinct misinformation patterns demand different verification strategies. Existing methods rely on either representation-driven classifiers or LLM-based retrieval-augmented and tool-using pipelines, which act as opaque black boxes or follow rigid, predefined tool chains, failing to adapt reasoning to diverse news and yielding weakly grounded, uninterpretable decisions. To this end, we propose MCTL, a Meta-Cognitive Tool-Learning Framework that reformulates detection from static classification into in-context cognitive orchestration with dynamically planned tool-based reasoning. Specifically, MCTL retrieves cross-modal contextual evidence as external memory to compensate for context sparsity, performs meta-cognitive planning to construct a sample-specific Directed Acyclic Graph (DAG) reasoning blueprint, routes specialized tools along the DAG, and fuses conflicting judgments via consensus-aware arbitration, enabling context-grounded, sample-adaptive, and traceable verification. Extensive experiments demonstrate that MCTL consistently outperforms recent state-of-the-art agentic and RAG baselines, confirming its effectiveness and reliability.

Retrieval heads, a subset of attention heads in Transformers, were studied in English, showing its crucial role in retrieving information from the context. We expand the study of retrieval heads to multilingual context and find that while nearly half of all retrieval heads are often shared across multiple languages, language-specific retrieval heads also emerge. We further design a cross-lingual needle-in-the-haystack task, to identify Retrieval Transition Heads (RTH) that retrieve key information and further map it to target-language output. Our experiments reveal that RTHs do not always overlap with RH, and can be more vital for Chain-of-Thought reasoning in multilingual LLMs. Across four multilingual benchmarks (MMLU-ProX, MGSM, MLQA, and XQuAD) and two model families (Qwen2.5-7B Instruct and Llama3.1-8B Instruct), we demonstrate that masking RTH induces bigger performance drop than masking Retrieval Heads (RH). Specifically, for Llama3.1-8B Instruct, masking the top-25 RTHs results in an average 37.3-point drop (against 19.5-point drop from RH masking) in reasoning accuracy (MMLU-ProX, MGSM) and 9.0 F1-score reduction (against 5.1 drop for RH) in extractive QA (MLQA, XQuAD). Our work advances the understanding of multilingual LMs by isolating the attention heads responsible for mapping to target languages. Code available at https://github.com/NerdyVisky/multilingual-retrieval-transition-heads

Token-level differentially private (DP) prompt optimization methods such as DP-OPT can become unstable under tight privacy budgets: on GSM8K, DP-OPT obtains 49.5 ± 28.5% across 30 runs, and a logged search trajectory reveals prompt-template drift and noise-sensitive irreversible choices. We diagnose these as structural consequences of greedy token-by-token construction over privately aggregated counts. We then propose DP-ES (Differentially Private Evolution Strategies), a structurally cleaner alternative that maintains a population of full prompts, mutates them via LLM calls that never access the private dataset, and spends privacy only on sampled-Gaussian evaluation; deterministic or Gumbel-smoothed selection is post-processing. Under a conservative (ε ≤ 1.0, δ = 10⁻⁵) guarantee, DP-ES achieves 88.1% on GSM8K (+38.6 pp over DP-OPT, approximately 9× lower standard deviation), 99.7% on MedQA, 73.5% on BANKING77, and 86.8% on Alpaca. It is also 2.5× faster in wall-clock time and uses 3.3× fewer logged private-data call groups than DP-OPT. Selection and population ablations, implementation-level noise checks, and a 200-profile exact-match memorization stress test complement the formal guarantee. Scope: Our experiments establish optimization robustness under DP noise, especially where prompt structure is critical; end-to-end validation on genuinely sensitive, non-saturated deployment data remains future work. Code: https://github.com/StephCpa/dp-es

Multilingual pretrained language models (mPLMs) have shown their effectiveness in multilingual word alignment induction. However, these methods usually start from mBERT or XLM-R. In this paper, we investigate whether multilingual sentence Transformer LaBSE is a strong multilingual word aligner. This idea is non-trivial as LaBSE is trained to learn language-agnostic sentence-level embeddings, while the alignment extraction task requires the more fine-grained word-level embeddings to be language-agnostic. We demonstrate that the vanilla LaBSE outperforms other mPLMs currently used in the alignment task, and then propose to finetune LaBSE on parallel corpus for further improvement. Experiment results on seven language pairs show that our best aligner outperforms previous state-of-the-art models of all varieties. In addition, our aligner supports different language pairs in a single model, and even achieves new state-of-the-art on zero-shot language pairs that does not appear in the finetuning process.