The paper presents an overview of the fourth edition of the Shared Task on Multilingual Coreference Resolution, organized as part of the CODI-CRAC 2025 workshop. As in the previous editions, participants were challenged to develop systems that identify mentions and cluster them according to identity coreference. A key innovation of this year's task was the introduction of a dedicated Large Language Model (LLM) track, featuring a simplified plaintext format designed to be more suitable for LLMs than the original CoNLL-U representation. The task also expanded its coverage with three new datasets in two additional languages, using version 1.3 of CorefUD - a harmonized multilingual collection of 22 datasets in 17 languages. In total, nine systems participated, including four LLM-based approaches (two fine-tuned and two using few-shot adaptation). While traditional systems still kept the lead, LLMs showed clear potential, suggesting they may soon challenge established approaches in future editions.

Training effective text rerankers is crucial for information retrieval. Two strategies are widely used: contrastive learning (optimizing directly on ground-truth labels) and knowledge distillation (transferring knowledge from a larger reranker). While both have been studied extensively, a clear comparison of their effectiveness for training cross-encoder rerankers under practical conditions is needed. This paper empirically compares these strategies by training rerankers of different sizes (0.5B, 1.5B, 3B, 7B) and architectures (Transformer, Recurrent) using both methods on the same data, with a strong contrastive learning model acting as the distillation teacher. Our results show that knowledge distillation generally yields better in-domain and out-of-domain ranking performance than contrastive learning when distilling from a more performant teacher model. This finding is consistent across student model sizes and architectures. However, distilling from a teacher of the same capacity does not provide the same advantage, particularly for out-of-domain tasks. These findings offer practical guidance for choosing a training strategy based on available teacher models. We recommend using knowledge distillation to train smaller rerankers if a larger, more performant teacher is accessible; in its absence, contrastive learning remains a robust baseline. Our code implementation is made available to facilitate reproducbility.

Sign Language Video Generation (SLVG) seeks to generate identity-preserving sign language videos from spoken language texts. Existing methods primarily rely on the single coarse condition (\eg, skeleton sequences) as the intermediary to bridge the translation model and the video generation model, which limits both the naturalness and expressiveness of the generated videos. To overcome these limitations, we propose SignViP, a novel SLVG framework that incorporates multiple fine-grained conditions for improved generation fidelity. Rather than directly translating error-prone high-dimensional conditions, SignViP adopts a discrete tokenization paradigm to integrate and represent fine-grained conditions (\ie, fine-grained poses and 3D hands). SignViP contains three core components. (1) Sign Video Diffusion Model is jointly trained with a multi-condition encoder to learn continuous embeddings that encapsulate fine-grained motion and appearance. (2) Finite Scalar Quantization (FSQ) Autoencoder is further trained to compress and quantize these embeddings into discrete tokens for compact representation of the conditions. (3) Multi-Condition Token Translator is trained to translate spoken language text to discrete multi-condition tokens. During inference, Multi-Condition Token Translator first translates the spoken language text into discrete multi-condition tokens. These tokens are then decoded to continuous embeddings by FSQ Autoencoder, which are subsequently injected into Sign Video Diffusion Model to guide video generation. Experimental results show that SignViP achieves state-of-the-art performance across metrics, including video quality, temporal coherence, and semantic fidelity. The code is available at https://github.com/umnooob/signvip/.

The recent success of generative AI highlights the crucial role of high-quality human feedback in building trustworthy AI systems. However, the increasing use of large language models (LLMs) by crowdsourcing workers poses a significant challenge: datasets intended to reflect human input may be compromised by LLM-generated responses. Existing LLM detection approaches often rely on high-dimensional training data such as text, making them unsuitable for annotation tasks like multiple-choice labeling. In this work, we investigate the potential of peer prediction -- a mechanism that evaluates the information within workers' responses without using ground truth -- to mitigate LLM-assisted cheating in crowdsourcing with a focus on annotation tasks. Our approach quantifies the correlations between worker answers while conditioning on (a subset of) LLM-generated labels available to the requester. Building on prior research, we propose a training-free scoring mechanism with theoretical guarantees under a crowdsourcing model that accounts for LLM collusion. We establish conditions under which our method is effective and empirically demonstrate its robustness in detecting low-effort cheating on real-world crowdsourcing datasets.

Properly evaluating the ability of Video-Language Models (VLMs) to understand long videos remains a challenge. We propose a long-context video understanding benchmark, Causal2Needles, that assesses two crucial abilities insufficiently addressed by existing benchmarks: (1) extracting information from two separate locations (two needles) in a long video and understanding them jointly, and (2) modeling the world in terms of cause and effect in human behaviors. Causal2Needles evaluates these abilities using noncausal one-needle, causal one-needle, and causal two-needle questions. The most complex question type, causal two-needle questions, require extracting information from both the cause and effect events from a long video and the associated narration text. To prevent textual bias, we introduce two complementary question formats: locating the video clip containing the answer, and verbal description of a visual detail from that video clip. Our experiments reveal that models excelling on existing benchmarks struggle with causal 2-needle questions, and the model performance is negatively correlated with the distance between the two needles. These findings highlight critical limitations in current VLMs. The dataset is available at: https://huggingface.co/datasets/causal2needles/Causal2Needles

We address the challenge of training diffusion models to sample from unnormalized energy distributions in the absence of data, the so-called diffusion samplers. Although these approaches have shown promise, they struggle to scale in more demanding scenarios where energy evaluations are expensive and the sampling space is high-dimensional. To address this limitation, we propose a scalable and sample-efficient framework that properly harmonizes the powerful classical sampling method and the diffusion sampler. Specifically, we utilize Monte Carlo Markov chain (MCMC) samplers with a novelty-based auxiliary energy as a Searcher to collect off-policy samples, using an auxiliary energy function to compensate for exploring modes the diffusion sampler rarely visits. These off-policy samples are then combined with on-policy data to train the diffusion sampler, thereby expanding its coverage of the energy landscape. Furthermore, we identify primacy bias, i.e., the preference of samplers for early experience during training, as the main cause of mode collapse during training, and introduce a periodic re-initialization trick to resolve this issue. Our method significantly improves sample efficiency on standard benchmarks for diffusion samplers and also excels at higher-dimensional problems and real-world molecular conformer generation.

Detecting subtle forms of dishonesty like sycophancy and manipulation in Large Language Models (LLMs) remains challenging for both humans and automated evaluators, as these behaviors often appear through small biases rather than clear false statements. We introduce Judge Using Safety-Steered Alternatives (JUSSA), a novel framework that employs steering vectors not to improve model behavior directly, but to enhance LLM judges' evaluation capabilities. JUSSA applies steering vectors during inference to generate more honest alternatives, providing judges with contrastive examples that make subtle dishonest patterns easier to detect. While existing evaluation methods rely on black-box evaluation, JUSSA leverages model internals to create targeted comparisons from single examples. We evaluate our method on sycophancy detection and introduce a new manipulation dataset covering multiple types of manipulation. Our results demonstrate that JUSSA effectively improves detection accuracy over single-response evaluation in various cases. Analysis across judge models reveals that JUSSA helps weaker judges on easier dishonesty detection tasks, and stronger judges on harder tasks. Layer-wise experiments show how dishonest prompts cause representations to diverge from honest ones in middle layers, revealing where steering interventions are most effective for generating contrastive examples. By demonstrating that steering vectors can enhance safety evaluation rather than just modify behavior, our work opens new directions for scalable model auditing as systems become increasingly sophisticated.

Generating high-quality MCQs, especially those targeting diverse cognitive levels and incorporating common misconceptions into distractor design, is time-consuming and expertise-intensive, making manual creation impractical at scale. Current automated approaches typically generate questions at lower cognitive levels and fail to incorporate domain-specific misconceptions. This paper presents a hierarchical concept map-based framework that provides structured knowledge to guide LLMs in generating MCQs with distractors. We chose high-school physics as our test domain and began by developing a hierarchical concept map covering major Physics topics and their interconnections with an efficient database design. Next, through an automated pipeline, topic-relevant sections of these concept maps are retrieved to serve as a structured context for the LLM to generate questions and distractors that specifically target common misconceptions. Lastly, an automated validation is completed to ensure that the generated MCQs meet the requirements provided. We evaluate our framework against two baseline approaches: a base LLM and a RAG-based generation. We conducted expert evaluations and student assessments of the generated MCQs. Expert evaluation shows that our method significantly outperforms the baseline approaches, achieving a success rate of 75.20% in meeting all quality criteria compared to approximately 37% for both baseline methods. Student assessment data reveal that our concept map-driven approach achieved a significantly lower guess success rate of 28.05% compared to 37.10% for the baselines, indicating a more effective assessment of conceptual understanding. The results demonstrate that our concept map-based approach enables robust assessment across cognitive levels and instant identification of conceptual gaps, facilitating faster feedback loops and targeted interventions at scale.

In many decision settings, the definitive ground truth is either non-existent or inaccessible. We introduce a framework for evaluating classifiers based solely on human judgments. In such cases, it is helpful to compare automated classifiers to human judgment. We quantify a classifier's performance by its rater equivalence: the smallest number of human raters whose combined judgment matches the classifier's performance. Our framework uses human-generated labels both to construct benchmark panels and to evaluate performance. We distinguish between two models of utility: one based on agreement with the assumed but inaccessible ground truth, and one based on matching individual human judgments. Using case studies and formal analysis, we demonstrate how this framework can inform the evaluation and deployment of AI systems in practice.

A longstanding goal in computational educational research is to develop explainable knowledge tracing (KT) models. Deep Knowledge Tracing (DKT), which leverages a Recurrent Neural Network (RNN) to predict student knowledge and performance on exercises, has been proposed as a major advancement over traditional KT methods. Several studies suggest that its performance gains stem from its ability to model bidirectional relationships between different knowledge components (KCs) within a course, enabling the inference of a student's understanding of one KC from their performance on others. In this paper, we challenge this prevailing explanation and demonstrate that DKT's strength lies in its implicit ability to model prerequisite relationships as a causal structure, rather than bidirectional relationships. By pruning exercise relation graphs into Directed Acyclic Graphs (DAGs) and training DKT on causal subsets of the Assistments dataset, we show that DKT's predictive capabilities align strongly with these causal structures. Furthermore, we propose an alternative method for extracting exercise relation DAGs using DKT's learned representations and provide empirical evidence supporting our claim. Our findings suggest that DKT's effectiveness is largely driven by its capacity to approximate causal dependencies between KCs rather than simple relational mappings.