Self-Consistent Latent Reasoning: Long Latent Sequence Reasoning for Vision-Language Model

In language reasoning, longer chains of thought consistently yield better performance, which naturally suggests that visual latent reasoning may likewise benefit from longer latent sequences. However, we discover a counterintuitive phenomenon: the performance of existing latent visual reasoning methods systematically degrades as the latent sequence grows longer. We reveal the root cause: Information Gain Collapse -- autoregressive generation makes each step highly dependent on prior outputs, so subsequent tokens can barely introduce new information. We further identify that heavily pooled ($\geq 128\times$) image embeddings used as supervision targets provide no more signal than meaningless placeholders. Motivated by these insights, we propose SCOLAR (Self-COnsistent LAtent Reasoning), which introduces a lightweight detransformer that leverages the LLM's full-sequence hidden states to generate auxiliary visual tokens in a single shot, with each token independently anchored to the original visual space. Combined with three-stage SFT and ALPO reinforcement learning, SCOLAR extends acceptable latent CoT length by over $30\times$, achieves state-of-the-art among open-source models on real-world reasoning benchmarks (+14.12% over backbone), and demonstrates strong out-of-distribution generalization.

StreamingClaw Technical Report

Applications such as embodied intelligence rely on a real-time perception-decision-action closed loop, posing stringent challenges for streaming video understanding. However, current agents suffer from fragmented capabilities, such as supporting only offline video understanding, lacking long-term multimodal memory mechanisms, or struggling to achieve real-time reasoning and proactive interaction under streaming inputs. These shortcomings have become a key bottleneck for preventing them from sustaining perception, making real-time decisions, and executing actions in real-world environments. To alleviate these issues, we propose StreamingClaw, a unified agent framework for streaming video understanding and embodied intelligence. It is also an OpenClaw-compatible framework that supports real-time, multimodal streaming interaction. StreamingClaw integrates five core capabilities: (1) It supports real-time streaming reasoning. (2) It supports reasoning about future events and proactive interaction under the online evolution of interaction objectives. (3) It supports multimodal long-term storage, hierarchical evolution, and efficient retrieval of shared memory across multiple agents. (4) It supports a closed-loop of perception-decision-action. In addition to conventional tools and skills, it also provides streaming tools and action-centric skills tailored for real-world physical environments. (5) It is compatible with the OpenClaw framework, allowing it to fully leverage the resources and support of the open-source community. With these designs, StreamingClaw integrates online real-time reasoning, multimodal long-term memory, and proactive interaction within a unified framework. Moreover, by translating decisions into executable actions, it enables direct control of the physical world, supporting practical deployment of embodied interaction.

ANPrompt: Anti-noise Prompt Tuning for Vision-Language Models

Prompt tuning has emerged as an efficient and effective technique for adapting vision-language models (VLMs) with low computational overhead. However, existing methods often overlook the vulnerability of prompt-tuned VLMs to weak semantic perturbations-such as subtle image or text noise-that degrade their generalization to unseen classes. To address this limitation, we propose ANPrompt, a novel prompt tuning framework designed to enhance robustness under such perturbations. ANPrompt first constructs weak noise text features by fusing original and noise-perturbed text embeddings, which are then clustered to form noise prompts. These noise prompts are integrated with learnable prompt tokens to generate anti-noise prompts, which are injected into the deeper layers of both image and text encoders. To further capture the noise-aware visual semantics, ANPrompt computes the Noise-Resistant Visual Prompt Prototype (NRVPP) by averaging the output prompt tokens from the vision encoder. Finally, ANPrompt introduces alignment, robustness, and anti-noise objectives by computing a Weak semantic noise Alignment Loss (WALoss) alongside the standard cross-entropy and sim loss. Experiments across 11 benchmarks demonstrate that ANPrompt consistently outperforms existing prompt tuning approaches, achieving superior robustness to semantic noise and improved generalization to novel categories.