Open-web future event prediction requires agents to distill reliable signals from noisy, redundant, and incomplete evidence. Existing retrieval/memory mechanisms directly feed retrieved information to agents or rely on simple memory functions such as storing and reusing prior information for prediction, leaving them insufficient for open-web forecasting. We propose to transform raw web evidence into structured memory before prediction, enabling agents to reason over distilled, question-specific evidence rather than noisy retrieval results. This paper presents ForeDreamer, a self-evolving dual-agent framework for managing memory over open-web evidence. ForeDreamer separates factual memory, a question-specific evidence state for the current forecast, from experiential memory, persistent agent experience accumulated across forecasting episodes. It uses a main agent for search and prediction, and a memory-processing subagent to convert search results into factual memory with dedicated tools. ForeDreamer further evolves experiential memory through two tracks, improving both forecasting decisions and factual-memory construction. Experiments on Prophet Arena and FutureX demonstrate the effectiveness of ForeDreamer.

Current Vision-Language Models struggle with hours-long videos because processing full-length visual sequences induces prohibitive token explosion and attention dilution. To overcome this, we introduce MemDreamer to decouple perception and reasoning, shifting long-video understanding into an agentic exploration process. As a plug-and-play framework, it incrementally streams videos to construct a Hierarchical Graph Memory, a top-down three-tier architecture for semantic abstraction, anchored by a foundational graph capturing spatiotemporal and causal relations. During inference, the reasoning model employs agentic tool-augmented retrieval, navigating hierarchies, searching nodes, and traversing logical edges via an Observation-Reason-Action loop. Experiments show MemDreamer achieves SOTA results across four mainstream benchmarks, substantially narrowing the gap with human experts. It greatly reduces the reasoning context relative to full-context ingestion while delivering a substantial absolute accuracy gain. Furthermore, statistical analysis uncovers a strong positive linear correlation between an VLM’s performance on logic reasoning and long-video understanding benchmarks, establishing agentic capability scaling as a new paradigm for multimodal comprehension.

Pretrained LLMs excel across diverse tasks, yet standard attention remains a major bottleneck for long-context understanding. Recent test-time training (TTT) methods alleviate this by encoding context into internal model parameters, but they typically require specialized architectures, meta-learned initialization, or continual post-training, preventing plug-and-play use on off-the-shelf LLMs. We propose test-time training sparse attention (TTT-Sparse), a plug-and-play alternative that performs test-time training over external memory while keeping the internal parameters frozen. For each input, TTT-Sparse constructs contextual chunk key states from the KV cache and learns a lightweight query adapter to induce a task-specific sparse attention pattern, which is used during generation to attend only to the most relevant context chunks and suppress interference from irrelevant context. Extensive experiments on long-context benchmarks show that TTT-Sparse improves long-context understanding while reducing generation-time attention cost.

Diffusion Language Models (DLMs) offer a promising alternative for language modeling by enabling parallel decoding through iterative refinement. However, most DLMs rely on hard binary masking and discrete token assignments, which hinder the revision of early decisions and underutilize intermediate probabilistic representations. In this paper, we propose EvoToken-DLM, a novel diffusion-based language modeling approach that replaces hard binary masks with evolving soft token distributions. EvoToken-DLM enables a progressive transition from masked states to discrete outputs, supporting revisable decoding. To effectively support this evolution, we introduce continuous trajectory supervision, which aligns training objectives with iterative probabilistic updates. Extensive experiments across multiple benchmarks show that EvoToken-DLM consistently achieves superior performance, outperforming strong diffusion-based and masked DLM baselines. Our code is available at https://github.com/aim-uofa/EvoTokenDLM.

Diffusion large language models (dLLMs) have recently attracted significant attention for their ability to enhance diversity, controllability, and parallelism. However, their non-sequential, bidirectionally masked generation makes quality assessment difficult, underscoring the need for effective self-evaluation. In this work, we propose DiSE, a simple yet effective self-evaluation confidence quantification method for dLLMs. DiSE quantifies confidence by computing the probability of regenerating the tokens in the entire generated sequence, given the full context. This method enables more efficient and reliable quality assessment by leveraging token regeneration probabilities, facilitating both likelihood estimation and robust uncertainty quantification. Building upon DiSE, we further introduce a flexible-length generation framework, which adaptively controls the sequence length based on the model’s self-assessment of its own output. We analyze and validate the feasibility of DiSE from the perspective of dLLM generalization, and empirically demonstrate that DiSE is positively correlated with both semantic coherence and answer accuracy. Extensive experiments on likelihood evaluation, uncertainty quantification, and flexible-length generation further confirm the effectiveness of the proposed DiSE.

Large Language Models (LLMs), such as LLaMA and T5, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LLMs on downstream tasks, their deployment is still hindered by the vast model scale and computational costs. Post-training model pruning offers a way to compress LLMs. However, the current pruning methods designed for LLMs are not compatible with LoRA. This is due to their utilization of unstructured pruning on LLMs, impeding the merging of LoRA weights, or their dependence on the gradients of pre-trained weights to guide pruning, which can impose significant memory overhead.To this end, we propose LoRAPrune, a new framework that delivers an accurate structured pruned model in a highly memory-efficient manner. Specifically, we first design a LoRA-guided pruning criterion, which uses the weights and gradients of LoRA, rather than the gradients of pre-trained weights for importance estimation. We subsequently integrate this criterion into an iterative pruning process, effectively removing redundant channels and heads. Extensive experimental results demonstrate the superior performance of our LoRAPrune over existing approaches on the LLaMA series models.At a 50% compression rate, LoRAPrune demonstrates superior performance over LLM-Pruner, achieving a reduction in perplexity by 4.81 on WikiText2 and 3.46 on PTB, while also decreasing memory usage by 52.6%.Besides, LoRAPrune also matches semi-structural pruning across multiple LLMs, proving its wide applicability. The code is available at https://github.com/aim-uofa/LoRAPrune.