Extending Memorization Dynamics in Pythia Models from Instance-Level Insights

Large language models have demonstrated a remarkable ability for verbatim memorization. While numerous works have explored factors influencing model memorization, the dynamic evolution memorization patterns remains underexplored. This paper presents a detailed analysis of memorization in the Pythia model family across varying scales and training steps under prefix perturbations. Using granular metrics, we examine how model architecture, data characteristics, and perturbations influence these patterns. Our findings reveal that: (1) as model scale increases, memorization expands incrementally while efficiency decreases rapidly; (2) as model scale increases, the rate of new memorization acquisition decreases while old memorization forgetting increases; (3) data characteristics (token frequency, repetition count, and uncertainty) differentially affect memorized versus non-memorized samples; and (4) prefix perturbations reduce memorization and increase generation uncertainty proportionally to perturbation strength, with low-redundancy samples showing higher vulnerability and larger models offering no additional robustness. These findings advance our understanding of memorization mechanisms, with direct implications for training optimization, privacy safeguards, and architectural improvements.

PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design

Designing protein-binding proteins with high affinity is critical in biomedical research and biotechnology. Despite recent advancements targeting specific proteins, the ability to create high-affinity binders for arbitrary protein targets on demand, without extensive rounds of wet-lab testing, remains a significant challenge. Here, we introduce PPDiff, a diffusion model to jointly design the sequence and structure of binders for arbitrary protein targets in a non-autoregressive manner. PPDiffbuilds upon our developed Sequence Structure Interleaving Network with Causal attention layers (SSINC), which integrates interleaved self-attention layers to capture global amino acid correlations, k-nearest neighbor (kNN) equivariant graph layers to model local interactions in three-dimensional (3D) space, and causal attention layers to simplify the intricate interdependencies within the protein sequence. To assess PPDiff, we curate PPBench, a general protein-protein complex dataset comprising 706,360 complexes from the Protein Data Bank (PDB). The model is pretrained on PPBenchand finetuned on two real-world applications: target-protein mini-binder complex design and antigen-antibody complex design. PPDiffconsistently surpasses baseline methods, achieving success rates of 50.00%, 23.16%, and 16.89% for the pretraining task and the two downstream applications, respectively.

Natural Language Guided Ligand-Binding Protein Design

Can AI protein models follow human language instructions and design proteins with desired functions (e.g. binding to a ligand)? Designing proteins that bind to a given ligand is crucial in a wide range of applications in biology and chemistry. Most prior AI models are trained on protein-ligand complex data, which is scarce due to the high cost and time requirements of laboratory experiments. In contrast, there is a substantial body of human-curated text descriptions about protein-ligand interactions and ligand formula. In this paper, we propose InstructPro, a family of protein generative models that follow natural language instructions to design ligand-binding proteins. Given a textual description of the desired function and a ligand formula in SMILES, InstructPro generates protein sequences that are functionally consistent with the specified instructions. We develop the model architecture, training strategy, and a large-scale dataset, InstructProBench, to support both training and evaluation. InstructProBench consists of 9,592,829 triples of (function description, ligand formula, protein sequence). We train two model variants: InstructPro-1B (with 1 billion parameters) and InstructPro-3B~(with 3 billion parameters). Both variants consistently outperform strong baselines, including ProGen2, ESM3, and Pinal. Notably, InstructPro-1B achieves the highest docking success rate (81.52% at moderate confidence) and the lowest average root mean square deviation (RMSD) compared to ground truth structures (4.026{\AA}). InstructPro-3B further descreases the average RMSD to 2.527{\AA}, demonstrating InstructPro's ability to generate ligand-binding proteins that align with the functional specifications.

HOI-PAGE: Zero-Shot Human-Object Interaction Generation with Part Affordance Guidance

We present HOI-PAGE, a new approach to synthesizing 4D human-object interactions (HOIs) from text prompts in a zero-shot fashion, driven by part-level affordance reasoning. In contrast to prior works that focus on global, whole body-object motion for 4D HOI synthesis, we observe that generating realistic and diverse HOIs requires a finer-grained understanding -- at the level of how human body parts engage with object parts. We thus introduce Part Affordance Graphs (PAGs), a structured HOI representation distilled from large language models (LLMs) that encodes fine-grained part information along with contact relations. We then use these PAGs to guide a three-stage synthesis: first, decomposing input 3D objects into geometric parts; then, generating reference HOI videos from text prompts, from which we extract part-based motion constraints; finally, optimizing for 4D HOI motion sequences that not only mimic the reference dynamics but also satisfy part-level contact constraints. Extensive experiments show that our approach is flexible and capable of generating complex multi-object or multi-person interaction sequences, with significantly improved realism and text alignment for zero-shot 4D HOI generation.

HardTests: Synthesizing High-Quality Test Cases for LLM Coding

Verifiers play a crucial role in large language model (LLM) reasoning, needed by post-training techniques such as reinforcement learning. However, reliable verifiers are hard to get for difficult coding problems, because a well-disguised wrong solution may only be detected by carefully human-written edge cases that are difficult to synthesize. To address this issue, we propose HARDTESTGEN, a pipeline for high-quality test synthesis using LLMs. With this pipeline, we curate a comprehensive competitive programming dataset HARDTESTS with 47k problems and synthetic high-quality tests. Compared with existing tests, HARDTESTGEN tests demonstrate precision that is 11.3 percentage points higher and recall that is 17.5 percentage points higher when evaluating LLM-generated code. For harder problems, the improvement in precision can be as large as 40 points. HARDTESTS also proves to be more effective for model training, measured by downstream code generation performance. We will open-source our dataset and synthesis pipeline at https://leililab.github.io/HardTests/.

SNS-Bench-VL: Benchmarking Multimodal Large Language Models in Social Networking Services

With the increasing integration of visual and textual content in Social Networking Services (SNS), evaluating the multimodal capabilities of Large Language Models (LLMs) is crucial for enhancing user experience, content understanding, and platform intelligence. Existing benchmarks primarily focus on text-centric tasks, lacking coverage of the multimodal contexts prevalent in modern SNS ecosystems. In this paper, we introduce SNS-Bench-VL, a comprehensive multimodal benchmark designed to assess the performance of Vision-Language LLMs in real-world social media scenarios. SNS-Bench-VL incorporates images and text across 8 multimodal tasks, including note comprehension, user engagement analysis, information retrieval, and personalized recommendation. It comprises 4,001 carefully curated multimodal question-answer pairs, covering single-choice, multiple-choice, and open-ended tasks. We evaluate over 25 state-of-the-art multimodal LLMs, analyzing their performance across tasks. Our findings highlight persistent challenges in multimodal social context comprehension. We hope SNS-Bench-VL will inspire future research towards robust, context-aware, and human-aligned multimodal intelligence for next-generation social networking services.

Wav2Sem: Plug-and-Play Audio Semantic Decoupling for 3D Speech-Driven Facial Animation

In 3D speech-driven facial animation generation, existing methods commonly employ pre-trained self-supervised audio models as encoders. However, due to the prevalence of phonetically similar syllables with distinct lip shapes in language, these near-homophone syllables tend to exhibit significant coupling in self-supervised audio feature spaces, leading to the averaging effect in subsequent lip motion generation. To address this issue, this paper proposes a plug-and-play semantic decorrelation module-Wav2Sem. This module extracts semantic features corresponding to the entire audio sequence, leveraging the added semantic information to decorrelate audio encodings within the feature space, thereby achieving more expressive audio features. Extensive experiments across multiple Speech-driven models indicate that the Wav2Sem module effectively decouples audio features, significantly alleviating the averaging effect of phonetically similar syllables in lip shape generation, thereby enhancing the precision and naturalness of facial animations. Our source code is available at https://github.com/wslh852/Wav2Sem.git.

Multimodal Mixture of Low-Rank Experts for Sentiment Analysis and Emotion Recognition

Multi-task learning (MTL) enables the efficient transfer of extra knowledge acquired from other tasks. The high correlation between multimodal sentiment analysis (MSA) and multimodal emotion recognition (MER) supports their joint training. However, existing methods primarily employ hard parameter sharing, ignoring parameter conflicts caused by complex task correlations. In this paper, we present a novel MTL method for MSA and MER, termed Multimodal Mixture of Low-Rank Experts (MMoLRE). MMoLRE utilizes shared and task-specific experts to distinctly model common and unique task characteristics, thereby avoiding parameter conflicts. Additionally, inspired by low-rank structures in the Mixture of Experts (MoE) framework, we design low-rank expert networks to reduce parameter and computational overhead as the number of experts increases. Extensive experiments on the CMU-MOSI and CMU-MOSEI benchmarks demonstrate that MMoLRE achieves state-of-the-art performance on the MSA task and competitive results on the MER task.

R2MED: A Benchmark for Reasoning-Driven Medical Retrieval

Current medical retrieval benchmarks primarily emphasize lexical or shallow semantic similarity, overlooking the reasoning-intensive demands that are central to clinical decision-making. In practice, physicians often retrieve authoritative medical evidence to support diagnostic hypotheses. Such evidence typically aligns with an inferred diagnosis rather than the surface form of a patient's symptoms, leading to low lexical or semantic overlap between queries and relevant documents. To address this gap, we introduce R2MED, the first benchmark explicitly designed for reasoning-driven medical retrieval. It comprises 876 queries spanning three tasks: Q&A reference retrieval, clinical evidence retrieval, and clinical case retrieval. These tasks are drawn from five representative medical scenarios and twelve body systems, capturing the complexity and diversity of real-world medical information needs. We evaluate 15 widely-used retrieval systems on R2MED and find that even the best model achieves only 31.4 nDCG@10, demonstrating the benchmark's difficulty. Classical re-ranking and generation-augmented retrieval methods offer only modest improvements. Although large reasoning models improve performance via intermediate inference generation, the best results still peak at 41.4 nDCG@10. These findings underscore a substantial gap between current retrieval techniques and the reasoning demands of real clinical tasks. We release R2MED as a challenging benchmark to foster the development of next-generation medical retrieval systems with enhanced reasoning capabilities. Data and code are available at https://github.com/R2MED/R2MED

Strategic Planning and Rationalizing on Trees Make LLMs Better Debaters

Winning competitive debates requires sophisticated reasoning and argument skills. There are unique challenges in the competitive debate: (1) The time constraints force debaters to make strategic choices about which points to pursue rather than covering all possible arguments; (2) The persuasiveness of the debate relies on the back-and-forth interaction between arguments, which a single final game status cannot evaluate. To address these challenges, we propose TreeDebater, a novel debate framework that excels in competitive debate. We introduce two tree structures: the Rehearsal Tree and Debate Flow Tree. The Rehearsal Tree anticipates the attack and defenses to evaluate the strength of the claim, while the Debate Flow Tree tracks the debate status to identify the active actions. TreeDebater allocates its time budget among candidate actions and uses the speech time controller and feedback from the simulated audience to revise its statement. The human evaluation on both the stage-level and the debate-level comparison shows that our TreeDebater outperforms the state-of-the-art multi-agent debate system. Further investigation shows that TreeDebater shows better strategies in limiting time to important debate actions, aligning with the strategies of human debate experts.