ChronoStitch: Training-Free Composition of Visual KV Memories for Long-Horizon Temporal Reasoning

• URL: http://arxiv.org/abs/2607.19547v1
• Date: Tue, 21 Jul 2026 19:46:38 GMT
• Title: ChronoStitch: Training-Free Composition of Visual KV Memories for Long-Horizon Temporal Reasoning
• Abstract summary: Long-video question answering requires a model to preserve visual evidence over time without repeatedly reprocessing the same video.<n>A practical approach is to store the vision-language model's internal key-value cache for each video chunk and retrieve that state at query time.<n>However, independently cached video chunks do not compose correctly.<n>This paper presents ChronoStitch, a training method for composing independently stored visual KV memories.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Long-video question answering requires a model to preserve visual evidence over time without repeatedly reprocessing the same video. A practical approach is to store the vision-language model's internal key-value (KV) cache for each video chunk and retrieve that state at query time. However, independently cached video chunks do not compose correctly: every chunk is prefilled from local rotary position zero, so naive concatenation collides temporal phases and removes the global order required for questions about what happened first, how often events occurred, or what changed across the video. This paper presents ChronoStitch, a training-free method for composing independently stored visual KV memories. The method first re-bases stored post-rotary keys onto a global three-axis multimodal RoPE coordinate system that preserves time, height, and width structure. We show why a one-dimensional scalar re-indexing is geometrically inconsistent for visual tokens because it turns spatial order within a frame into false temporal displacement. We then address the residual content gap left by positional repair: later chunks were originally encoded without attending to earlier chunks. ChronoStitch therefore selectively recomputes a small fraction of high-deviation later-chunk visual tokens while allowing them to attend over the composed cache. On Qwen2.5-VL-3B and the temporal split of TempCompass, ChronoStitch outperforms naive composition and position-only variants, improving event-ordering accuracy while running 3.3x faster than full joint re-prefilling.

Related papers

• Addressable Memory for Video World Models [50.40232922140905]
  We study visual persistence in interactive video world models.<n>We propose WorldTrace, a training-free memory framework for long-horizon visual persistence.
  arXiv  Detail & Related papers  (2026-08-07T16:55:57Z)
• Closing the Loop: Training-Free Revisit Consistency for Autoregressive Generative Rendering [8.16620085872554]
  New conditional video generation models have shown promising potentials to transform 3D engine renderings into photorealistic videos for gaming and immersive content creation.<n>These applications require long-horizon auto-regressive generation that continuously synthesizes new frames while preserving a persistent 3D world.<n>We demonstrate our method on loop-closure trajectories mined from TartanAir and Tartan dataset to mirror complicate real-world application scenarios.
  arXiv  Detail & Related papers  (2026-07-23T22:33:52Z)
• Position Rebinding Cache Reuse: Replay-Free Visual Revisiting for Interleaved Multimodal Reasoning [51.563653495547335]
  multimodal reasoning improves visual grounding by revisiting visual evidence during multi-step generation.<n>We propose Position Rebinding Cache Reuse (PRCR), a cache-level framework for replay-free visual revisiting.<n>PRCR achieves replay-level or better performance, improving average accuracy by 5 percent and reducing visual-revisiting by up to tens of thousands of times.
  arXiv  Detail & Related papers  (2026-06-25T05:47:52Z)
• Kamera: Unified Position-Invariant Multimodal KV Cache for Training-Free Reuse [1.081571058570587]
  Multimodal agents repeatedly re-examine the same video frames, UI screenshots, and rendered artifacts as their context window slides.<n>We show this recompute is avoidable, and identify exactly what naive KV reuse loses.<n>Blind reuse therefore leaves single-hop recall intact while halving multi-hop accuracy.<n>We repair it with a small, training-free low-rank conditioning patch stored alongside each position-free chunk.
  arXiv  Detail & Related papers  (2026-06-22T16:47:00Z)
• DynaTok: Token-Based 4D Reconstruction from Partial Point Clouds [86.5916611076068]
  DynaTok is a point-based framework for correspondence-free 4D reconstruction from partial point sequences without images.<n>A flow-matching decoder then reconstructs complete, temporally consistent 4D point-temporal observations.
  arXiv  Detail & Related papers  (2026-06-10T15:13:14Z)
• FadeMem: Distance-Aware Memory Consolidation for Autoregressive Video Diffusion [59.207505503284715]
  FadeMem organizes historical KV blocks into a temporal hierarchy under a fixed cache budget.<n>New history is inserted as fine-grained entries, while older adjacent entries are progressively merged.<n>Experiments show improved subject consistency, background stability, and temporal coherence over existing bounded-cache strategies.
  arXiv  Detail & Related papers  (2026-06-09T10:22:18Z)
• WorldKV: Efficient World Memory with World Retrieval and Compression [56.57011243315561]
  We propose WorldKV, a training-free framework with two components: World Retrieval and World Compression.<n>On Matrix-Game-2.0 and LingBot- World-Fast, WorldKV matches or exceeds full- KV memory fidelity at roughly 2x the throughput, and is competitive with memory-trained baselines without any fine-tuning.
  arXiv  Detail & Related papers  (2026-05-21T16:55:04Z)
• TempoFit: Plug-and-Play Layer-Wise Temporal KV Memory for Long-Horizon Vision-Language-Action Manipulation [16.28432866472846]
  We present TempoFit, a training-free temporal retrofit that upgrades frozen Vision-Language-Action policies through state-level memory.<n>Our key insight is that prefix attention K/V already form a model-native, content-addressable runtime state.<n>On LIBERO-LONG, TempoFit improves strong pretrained backbones by up to +4.0% average success rate while maintaining near-real-time latency.
  arXiv  Detail & Related papers  (2026-03-08T14:17:25Z)
• PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cache [61.57938553036056]
  We introduce PackCache, a training-free KV-cache management method that compacts the KV cache through three coordinated mechanisms.<n>In terms of efficiency, PackCache accelerates end-to-end generation by 1.7-2.2x on 48-frame long sequences.
  arXiv  Detail & Related papers  (2026-01-07T19:51:06Z)
• Temporally Consistent Transformers for Video Generation [80.45230642225913]
  To generate accurate videos, algorithms have to understand the spatial and temporal dependencies in the world. No established benchmarks on complex data exist for rigorously evaluating video generation with long temporal dependencies. We introduce the Temporally Consistent Transformer (TECO), a generative model that substantially improves long-term consistency while also reducing sampling time.
  arXiv  Detail & Related papers  (2022-10-05T17:15:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.