Ask Anything is an open source project that lets you chat with a video and ask questions about what happens in it. It is built by OpenGVLab and released under the MIT license. You can find the code and demos at OpenGVLab/Ask-Anything t.co
3 views
Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction
Video event prediction (VEP) requires models to infer unobserved future states from partial video evidence. Existing video MLLMs usually verbalize intermediate future reasoning in text space: once visual evidence is verbalized, fine-grained motion, geometry, and interaction cues can be lost, leading to plausible but visually ungrounded hallucinations. We introduce Future-L1, an interleaved latent visual reasoning framework that lets an MLLM alternate between language tokens and continuous latent visual spans during autoregressive decoding. To train this capability, we construct Future-L1-50K by selecting examples where future visual hints help prediction and align latent states to future-frame embeddings, then further optimize sampled latent trajectories with LA-DAPO, a latent-aware RL objective with outcome-contrastive and temporal-diversity rewards. Future-L1 achieves new state-of-the-art results on both benchmarks: on FutureBench, it improves Qwen3-VL-8B from 61.0 to 85.4 and exceeds the previous best Video-CoE by 10.4 points; on TwiFF-Bench, it improves the average score from 2.44 to 3.04. These results suggest that future-oriented video reasoning benefits from preserving intermediate visual semantics in latent space rather than translating every reasoning step into text.
Comments: Accepted to EMNLP 2026 Main Conference. Code: OpenGVLab/Future-L1
🚨 Thrilled to introduce RSI-Nav: Data-Centric RSI for Navigation!
Built on Qwen3.5-9B, it unifies VL2A navigation, VA2L instruction generation & spatial understanding, then evolves through its own verified data. 🧵👇
📄 arxiv.org
💻 OpenGVLab/RSI-Nav t.co
2 likes · 1 repost · 57 views
Ask Anything: Chat With Any Video
Ask Anything is an open source project that lets you chat with a video and ask questions about what happens in it. It is built by OpenGVLab and released under the MIT license. You can find the code and demos at OpenGVLab/Ask-Anything
8 likes · 1 comment · 922 views
能聊天的 AI 到处都是，能看图还动手的不多——InternGPT 让你指着图跟它说话 🖱️
OpenGVLab/InternGPT
6 views
Mentions
Ask Anything is an open source project that lets you chat with a video and ask questions about what happens in it. It is built by OpenGVLab and released under the MIT license. You can find the code and demos at OpenGVLab/Ask-Anything t.co
Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction Video event prediction (VEP) requires models to infer unobserved future states from partial video evidence. Existing video MLLMs usually verbalize intermediate future reasoning in text space: once visual evidence is verbalized, fine-grained motion, geometry, and interaction cues can be lost, leading to plausible but visually ungrounded hallucinations. We introduce Future-L1, an interleaved latent visual reasoning framework that lets an MLLM alternate between language tokens and continuous latent visual spans during autoregressive decoding. To train this capability, we construct Future-L1-50K by selecting examples where future visual hints help prediction and align latent states to future-frame embeddings, then further optimize sampled latent trajectories with LA-DAPO, a latent-aware RL objective with outcome-contrastive and temporal-diversity rewards. Future-L1 achieves new state-of-the-art results on both benchmarks: on FutureBench, it improves Qwen3-VL-8B from 61.0 to 85.4 and exceeds the previous best Video-CoE by 10.4 points; on TwiFF-Bench, it improves the average score from 2.44 to 3.04. These results suggest that future-oriented video reasoning benefits from preserving intermediate visual semantics in latent space rather than translating every reasoning step into text. Comments: Accepted to EMNLP 2026 Main Conference. Code: OpenGVLab/Future-L1
🚨 Thrilled to introduce RSI-Nav: Data-Centric RSI for Navigation! Built on Qwen3.5-9B, it unifies VL2A navigation, VA2L instruction generation & spatial understanding, then evolves through its own verified data. 🧵👇 📄 arxiv.org 💻 OpenGVLab/RSI-Nav t.co
Ask Anything: Chat With Any Video Ask Anything is an open source project that lets you chat with a video and ask questions about what happens in it. It is built by OpenGVLab and released under the MIT license. You can find the code and demos at OpenGVLab/Ask-Anything
能聊天的 AI 到处都是，能看图还动手的不多——InternGPT 让你指着图跟它说话 🖱️ OpenGVLab/InternGPT