PMMC: Prospective Multimodal Memory Compilation for Long-Term LVLM Agents

Long-term memory is essential for LVLM agents to maintain consistency and integrate information across extended multimodal interactions. Existing agent memory systems, however, often reduce visual experiences into textual summaries or rely on static retrieve-then-reason pipelines, which are inefficient at query time and brittle when questions require image-text binding, temporal updates, or visual details. We propose Prospective Multimodal Memory Compilation, a framework that shifts part of the memory reasoning process from query time to memory consolidation time. Given accumulated multimodal interactions, a Questioner predicts future question candidates, a Planner compiles question-conditioned multimodal memory programs, and a Doubter verifies whether the planned evidence path can support the predicted answer. The verified question-program pairs form a structured question bank for efficient query-time routing and evidence retrieval. Experiments on multimodal long-term memory benchmarks show that our method improves answer quality and visual evidence recall while reducing query-time token and latency costs. Extensive ablations analyze the effects of self-feedback, dynamic planning, raw-image access, and question bank coverage.

TrajWiki: Source-Grounded Memory Trajectories for Long-Horizon Dialogue Agents

Large language model agents have shown strong capabilities in generating coherent and contextually appropriate responses, yet robust long-horizon dialogue remains limited by the lack of external memory that is traceable, updatable, and diagnostically transparent. Existing memory-augmented agents often store memories as isolated records or overwritable states, making it difficult to preserve how information originates, evolves, conflicts, or becomes obsolete over time. We propose TrajWiki, a trajectory-based memory framework for long-horizon conversational agents. Instead of treating memory as static entries, TrajWiki represents each memory as a source-grounded evolution trajectory, maintained through immutable episodic snapshots and claim-level operations such as ADD, REVISE, and DEPRECATE. To reduce fragmentation and retrieval cost, TrajWiki further introduces Memory Wiki, a persistent intermediate layer that incrementally compiles dialogue history into structured and interlinked wiki pages capturing salient entities, events, quantities, topics, and conflicts. At inference time, queries are routed hierarchically from relevant wiki pages to linked memory trajectories, then to corresponding snapshots and source messages for evidence-grounded answer synthesis. Experiments on LoCoMo and MedMT show that TrajWiki improves long-horizon dialogue performance across both open-source and closed-source LLM backbones, while providing greater interpretability and diagnostic visibility into memory evolution, retrieval failures, and answer generation.

Stress Testing Concept Erasure with Large Language Model Agents

Concept erasure aims to remove semantic concepts from a trained generative model and is increasingly important for responsible AI deployment. However, verifying whether a model has robustly removed targeted concepts remains a critical challenge. Existing evaluation methods are typically pre-defined and static, failing to expose vulnerabilities under diverse natural-language probes and challenging conditions. Moreover, manually designed evaluation strategies can be biased and difficult to scale. We posit that concept erasure evaluation is best formulated as an adaptive hypothesis search, operationalised by agents that iteratively propose, critique, and verify tests to systematically expand coverage of failure modes. To this end, we propose Stress Testing Agents for Concept Erasure (STACE), a framework that autonomously stress-tests concept-erased models using multiple Large Language Model (LLM) agents, by iteratively generating and verifying stress-testing hypotheses grounded by external knowledge. We also introduce a suite of metrics for assessing the performance and efficiency of LLM-agent-powered stress-testing frameworks. Our extensive experiments show that STACE outperforms five LLM-based evaluation baselines on four concept categories. Further analysis across two T2I models, six concept erasure approaches, and various erasure strengths show that STACE is robust for different settings. We also show that STACE can be adapted beyond concept erasure evaluation to other problem domains, such as LLM jailbreaking. Our code is available anonymously.

Are Reasoning Vision-Language Models Robust to Semantic Visual Distractions?

Reasoning Vision-Language Models (VLMs) achieve strong performance on complex multimodal tasks, but reliable real-world application requires handling visual inputs that are messier than clean, curated benchmarks. Existing works mainly evaluate such reliability of VLMs through input corruptions, such as noise, blur and weather effects, which make visual evidence harder to perceive. This leaves a critical reliability failure mode underexplored: a model may perceive the evidence correctly, yet reason from plausible but irrelevant and distracting evidence and propagate this mistake to its final answer. To address this gap, we introduce \textbf{Distract-Bench}, a benchmark for evaluating VLM robustness to \textbf{semantic visual distractions}, defined as meaningful but task-irrelevant visual cues added to inputs while preserving the ground-truth answer. We comprehensively evaluate eight leading open-source and two closed-source VLMs across conventional vision corruptions and Distract-Bench. Our results show that Distract-Bench exposes a robustness failure distinct from vision corruptions: reasoning VLMs largely track their non-reasoning base models under perceptual degradation, but show consistently lower robustness to semantic distractions. Further analysis shows that these distractions often enter the reasoning process of VLMs, are treated as evidence, and lead to incorrect answers. Together, these findings reframe robustness evaluation for reasoning VLMs, shifting the focus from degraded perception to distractions for reliable real-world visual reasoning. Our data and code are available at https://github.com/Yizheng-Sun/Distract-Bench.

MultiwayPAM: Multiway Partitioning Around Medoids for LLM-as-a-Judge Score Analysis

LLM-as-a-Judge is a flexible framework for text evaluation, which allows us to obtain scores for the quality of a given text from various perspectives by changing the prompt template. Two main challenges in using LLM-as-a-Judge are computational cost of LLM inference, especially when evaluating a large number of texts, and inherent bias of an LLM evaluator. To address these issues and reveal the structure of score bias caused by an LLM evaluator, we propose to apply a tensor clustering method to a given LLM-as-a-Judge score tensor, whose entries are the scores for different combinations of questions, answerers, and evaluators. Specifically, we develop a new tensor clustering method MultiwayPAM, with which we can simultaneously estimate the cluster membership and the medoids for each mode of a given data tensor. By observing the medoids obtained by MultiwayPAM, we can gain knowledge about the membership of each question/answerer/evaluator cluster. We experimentally show the effectiveness of MultiwayPAM by applying it to the score tensors for two practical datasets.

Empirical Cumulative Distribution Function Clustering for LLM-based Agent System Analysis

Large language models (LLMs) are increasingly used as agents to solve complex tasks such as question answering (QA), scientific debate, and software development. A standard evaluation procedure aggregates multiple responses from LLM agents into a single final answer, often via majority voting, and compares it against reference answers. However, this process can obscure the quality and distributional characteristics of the original responses. In this paper, we propose a novel evaluation framework based on the empirical cumulative distribution function (ECDF) of cosine similarities between generated responses and reference answers. This enables a more nuanced assessment of response quality beyond exact match metrics. To analyze the response distributions across different agent configurations, we further introduce a clustering method for ECDFs using their distances and the $k$-medoids algorithm. Our experiments on a QA dataset demonstrate that ECDFs can distinguish between agent settings with similar final accuracies but different quality distributions. The clustering analysis also reveals interpretable group structures in the responses, offering insights into the impact of temperature, persona, and question topics.

Seed1.5-VL Technical Report

We present Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning. Seed1.5-VL is composed with a 532M-parameter vision encoder and a Mixture-of-Experts (MoE) LLM of 20B active parameters. Despite its relatively compact architecture, it delivers strong performance across a wide spectrum of public VLM benchmarks and internal evaluation suites, achieving the state-of-the-art performance on 38 out of 60 public benchmarks. Moreover, in agent-centric tasks such as GUI control and gameplay, Seed1.5-VL outperforms leading multimodal systems, including OpenAI CUA and Claude 3.7. Beyond visual and video understanding, it also demonstrates strong reasoning abilities, making it particularly effective for multimodal reasoning challenges such as visual puzzles. We believe these capabilities will empower broader applications across diverse tasks. In this report, we mainly provide a comprehensive review of our experiences in building Seed1.5-VL across model design, data construction, and training at various stages, hoping that this report can inspire further research. Seed1.5-VL is now accessible at https://www.volcengine.com/ (Volcano Engine Model ID: doubao-1-5-thinking-vision-pro-250428)

Dynamic Latency for CTC-Based Streaming Automatic Speech Recognition With Emformer

An inferior performance of the streaming automatic speech recognition models versus non-streaming model is frequently seen due to the absence of future context. In order to improve the performance of the streaming model and reduce the computational complexity, a frame-level model using efficient augment memory transformer block and dynamic latency training method is employed for streaming automatic speech recognition in this paper. The long-range history context is stored into the augment memory bank as a complement to the limited history context used in the encoder. Key and value are cached by a cache mechanism and reused for next chunk to reduce computation. Afterwards, a dynamic latency training method is proposed to obtain better performance and support low and high latency inference simultaneously. Our experiments are conducted on benchmark 960h LibriSpeech data set. With an average latency of 640ms, our model achieves a relative WER reduction of 6.0% on test-clean and 3.0% on test-other versus the truncate chunk-wise Transformer.

Locality Matters: A Locality-Biased Linear Attention for Automatic Speech Recognition

Conformer has shown a great success in automatic speech recognition (ASR) on many public benchmarks. One of its crucial drawbacks is the quadratic time-space complexity with respect to the input sequence length, which prohibits the model to scale-up as well as process longer input audio sequences. To solve this issue, numerous linear attention methods have been proposed. However, these methods often have limited performance on ASR as they treat tokens equally in modeling, neglecting the fact that the neighbouring tokens are often more connected than the distanced tokens. In this paper, we take this fact into account and propose a new locality-biased linear attention for Conformer. It not only achieves higher accuracy than the vanilla Conformer, but also enjoys linear space-time computational complexity. To be specific, we replace the softmax attention with a locality-biased linear attention (LBLA) mechanism in Conformer blocks. The LBLA contains a kernel function to ensure the linear complexities and a cosine reweighing matrix to impose more weights on neighbouring tokens. Extensive experiments on the LibriSpeech corpus show that by introducing this locality bias to the Conformer, our method achieves a lower word error rate with more than 22% inference speed.

HMM-Free Encoder Pre-Training for Streaming RNN Transducer

This work describes an encoder pre-training procedure using frame-wise label to improve the training of streaming recurrent neural network transducer (RNN-T) model. Streaming RNN-T trained from scratch usually performs worse and has high latency. Although it is common to address these issues through pre-training components of RNN-T with other criteria or frame-wise alignment guidance, the alignment is not easily available in end-to-end manner. In this work, frame-wise alignment, used to pre-train streaming RNN-T's encoder, is generated without using a HMM-based system. Therefore an all-neural framework equipping HMM-free encoder pre-training is constructed. This is achieved by expanding the spikes of CTC model to their left/right blank frames, and two expanding strategies are proposed. To our best knowledge, this is the first work to simulate HMM-based frame-wise label using CTC model. Experiments conducted on LibriSpeech and MLS English tasks show the proposed pre-training procedure, compared with random initialization, reduces the WER by relatively 5%~11% and the emission latency by 60 ms. Besides, the method is lexicon-free, so it is friendly to new languages without manually designed lexicon.