Beyond Final Answers: CRYSTAL Benchmark for Transparent Multimodal Reasoning Evaluation

We introduce CRYSTAL (Clear Reasoning via Yielded Steps, Traceability, and Logic), a diagnostic benchmark with 6,372 instances that evaluates multimodal reasoning through verifiable intermediate steps. We propose two complementary metrics: Match F1, which scores step-level precision and recall via semantic similarity matching, and Ordered Match F1, which further penalizes disordered reasoning chains. References are constructed through a Delphi-inspired pipeline in which four independent MLLMs generate trajectories, which are then aggregated via semantic clustering and validated through human quality gates. Evaluation of 20 MLLMs, including commercial frontier systems not used during benchmark construction, reveals systematic failures that are invisible to answer accuracy: universal cherry-picking (precision far exceeds recall), non-monotonic scaling trade-offs, and disordered reasoning in which no competitive model preserves more than 60% of matched steps in the correct order. Beyond evaluation, we propose the Causal Process Reward (CPR), a multiplicative reward that couples answer correctness with step-level alignment, and CPR-Curriculum, which progressively increases reasoning difficulty during training. CPR-Curriculum achieves a 32% improvement in Match F1 via GRPO where additive reward strategies fail, improving reasoning without manual step annotation.

Native LLM and MLLM Inference at Scale on Apple Silicon

The growing adoption of Apple Silicon for machine learning development has created demand for efficient inference solutions that leverage its unique unified memory architecture. However, existing tools either lack native optimization (PyTorch MPS) or focus solely on text models (llama.cpp), leaving multimodal workloads underserved. We present vllm-mlx, a framework for efficient LLM and MLLM inference on Apple Silicon built natively on MLX. For text models, we achieve 21% to 87% higher throughput than llama.cpp across models ranging from Qwen3-0.6B to Nemotron-30B, while providing continuous batching that scales to 4.3x aggregate throughput at 16 concurrent requests. For multimodal models, we introduce content-based prefix caching that eliminates redundant vision encoding by identifying identical images through content hashing, regardless of input format. Our evaluation on Apple M4 Max demonstrates throughput of up to 525 tokens per second on text models and 28x speedup on repeated image queries, reducing multimodal latency from 21.7 seconds to under 1 second. Video analysis with up to 64 frames achieves 24.7x cache speedup. We release our implementation as open source to support efficient inference on consumer Apple hardware.

Multi-layer Learnable Attention Mask for Multimodal Tasks

While the Self-Attention mechanism in the Transformer model has proven to be effective in many domains, we observe that it is less effective in more diverse settings (e.g. multimodality) due to the varying granularity of each token and the high computational demands of lengthy sequences. To address the challenges, we introduce the Learnable Attention Mask (LAM), strategically designed to globally regulate attention maps and prioritize critical tokens within the sequence. Leveraging the Self-Attention module in a BERT-like transformer network, our approach adeptly captures associations between tokens. The extension of the LAM to a multi-layer version accommodates the varied information aspects embedded at each layer of the Transformer network. Comprehensive experimental validation on various datasets, such as MADv2, QVHighlights, ImageNet 1K, and MSRVTT, demonstrates the efficacy of the LAM, exemplifying its ability to enhance model performance while mitigating redundant computations. This pioneering approach presents a significant advancement in enhancing the understanding of complex scenarios, such as in movie understanding.

FT2TF: First-Person Statement Text-To-Talking Face Generation

Talking face generation has gained immense popularity in the computer vision community, with various applications including AR/VR, teleconferencing, digital assistants, and avatars. Traditional methods are mainly audio-driven ones which have to deal with the inevitable resource-intensive nature of audio storage and processing. To address such a challenge, we propose FT2TF - First-Person Statement Text-To-Talking Face Generation, a novel one-stage end-to-end pipeline for talking face generation driven by first-person statement text. Moreover, FT2TF implements accurate manipulation of the facial expressions by altering the corresponding input text. Different from previous work, our model only leverages visual and textual information without any other sources (e.g. audio/landmark/pose) during inference. Extensive experiments are conducted on LRS2 and LRS3 datasets, and results on multi-dimensional evaluation metrics are reported. Both quantitative and qualitative results showcase that FT2TF outperforms existing relevant methods and reaches the state-of-the-art. This achievement highlights our model capability to bridge first-person statements and dynamic face generation, providing insightful guidance for future work.

Localizing Moments in Long Video Via Multimodal Guidance

The recent introduction of the large-scale long-form MAD dataset for language grounding in videos has enabled researchers to investigate the performance of current state-of-the-art methods in the long-form setup, with unexpected findings. In fact, current grounding methods alone fail at tackling this challenging task and setup due to their inability to process long video sequences. In this work, we propose an effective way to circumvent the long-form burden by introducing a new component to grounding pipelines: a Guidance model. The purpose of the Guidance model is to efficiently remove irrelevant video segments from the search space of grounding methods by coarsely aligning the sentence to chunks of the movies and then applying legacy grounding methods where high correlation is found. We term these video segments as non-describable moments. This two-stage approach reveals to be effective in boosting the performance of several different grounding baselines on the challenging MAD dataset, achieving new state-of-the-art performance.