cs.CVNov 11, 2026

Frame First, Decode Once: Object-Centric Coordinates and Single-Pass Decoding for System-1 World Models

Authors: Abu Anas Ibn Samad

Organizations: Sigmind.ai, Dhaka 1215, Bangladesh

Abstract

Single-pass primitive decoders such as LocateAnything and VFlash predict all quantized coordinate tokens of a geometric primitive in one forward pass, enabling fast, System-1 visual planning. We study what matters when such a decoder serves as a System-1 world model: a fast action-conditioned predictor that ranks candidate pushes by their predicted outcome. On ManiSkill3 PushT-v1 (512 held-out test families, 8,192 rollouts) and on human demonstrations from lerobot/pusht (2,384 test windows), we compare decoders, training losses, rigid-body projections and coordinate frames, using matched checkpoints, hierarchical bootstrap confidence intervals and Holm correction. Two factors of similar size dominate. First, expressing inputs and outputs in the SE(2) frame of the object's initial pose lowers decision regret by 1.9-2.2x for each of five decoders, and the object-frame model trained on 128 families matches the world-frame model trained on 512. Second, single-pass decoding outperforms iterative decoding: in the object frame, single-pass heads reach 0.020-0.021 regret at 1.6-1.8 ms per decision, against 0.046 for 10-step DDIM diffusion and 0.059 for KV-cached autoregression at 10.3 and 12.9 ms. The choice among single-pass heads has no measurable effect on regret, and our registered hypothesis that the frame effect exceeds the decoder spread is not supported. Soft-token expectation with an end-to-end 2D Kabsch projection gives the lowest trajectory error (11.35 px on ManiSkill3, 3.78 px on lerobot/pusht), exact rigid-body consistency, and lower regret than diffusion at every budget from 4 to 50 DDIM steps, at 3-29x lower latency.

Figures & tables

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Oct 2, 2026

Keeping JEPA World Models Plannable When Little of the Frame Moves

Specifying a goal in language rather than as a goal frame is a natural interface for planning with a latent world model, but testing it needs scenes in which language must discriminate between several objects. We build SLIM, a pushing benchmark with several small objects and paired visual and language goals on identical scenes. On SLIM a LeWM world model that solves PushT succeeds on under 1% of trials, although a scripted controller with simulator state solves every tier. Probes locate the failure in the encoder: its latent is nearly action-insensitive, neither pusher nor object positions can be decoded from it, and rollouts are no better than copying the current latent forward. One inverse-dynamics auxiliary loss, applied to encoder latents and to predicted latents through a shared head discarded at test time, restores every probe and raises success from 0.003 to 0.35 (0.16 on the hard pushing tier, where a goal-agnostic policy scores zero), and improves PushT at twice the trained horizon. Controls attribute the repair to the gradient into the encoder, and a response sweep shows that the vanilla model plans once enough of the frame responds to actions. A cheap action-sensitivity probe, computable without environment access, acts as an empirical necessary condition: all configurations below its threshold failed to plan. On the repaired latent, a small language-goal head plans from sentences without retraining the world model: it reaches 0.84 on navigation (visual-goal oracle 1.00), follows the named zone when it is swapped with a decoy, and degrades gracefully to unseen nouns. A single goal sentence rarely completes a push, but given the push as a sequence of stage sentences the head raises success on the medium and hard pushing tiers from 0.04 to 0.25, on par with the goal-frame oracle, also when the switch between stages is read from the latent alone.
Oct 26, 2026

LOCUS: Locate Once, Check Until Settled. Draft-and-Verify Visual Grounding with a Box Head and Coordinate-Token Verification

Vision-language models (VLMs) ground a referring expression in one of two ways. They decode the box as coordinate tokens, which is precise but takes one sequential step per token, or they regress it from a hidden state with a small head, which takes one step but loses accuracy at high IoU. The two fail differently: a head usually finds the right object but places its edges loosely, while coordinate tokens place edges precisely. We therefore use the regression head as a speculative drafter for the model's own coordinate tokens. Locus writes the head's box as coordinate tokens and checks them with the same backbone by Jacobi iteration over a shared image prefix. The fixed point is the model's greedy coordinate-token answer, reached in about 5.60 sequential passes on average instead of 18.46 decoding steps; head and verifier are one set of weights, trained jointly. On the 8 RefCOCO, RefCOCO+ and RefCOCOg splits, verification raises Locus-2B's [email protected] from 54.87 for its own head to 59.13, and the one-pass head is ahead at [email protected] of every published model with at most 3B parameters except Florence-2-L. Against Jag, the strongest open regression-head model, at its own size and training budget, Locus ties at [email protected] and, with typical acceptance, gains +4.52 points at [email protected], with paired significance tests. Exact answers come 1.5x faster than vLLM greedy-decoding the same coordinate tokens with the same weights; a pure regression head remains faster still.
Oct 11, 2026

1Step-WAM: Your World Action Model but Faster using 1-Step Action Generation

World Action Models (WAMs) unify visual world modeling and action generation, yet their real-world deployment is heavily bottlenecked by iterative action denoising and the sequential execution of visual prefilling and action prediction. We demonstrate that the action-denoising trajectories in WAMs are approximately straight, rendering the first-step velocity a strong estimate of the final action direction. Building on this observation, we introduce 1Step-WAM, a framework that transforms multi-step action denoising into one single forward. Specifically, we employ a lightweight calibrator to directly correct the first step velocity, and use bandwidth-constrained action decoding to further reduce errors in the calibrated one-step actions. Compressing action generation into a single step enables it to execute in parallel with visual prefilling, completely breaking the sequential bottleneck of the traditional imagination-then-action paradigm. Evaluations on RoboTwin and LIBERO show that 1Step-WAM largely preserves multi-step task success with only about 13.4M trainable parameters, while achieving inference speedups of 7.44×7.44\times, 5.98×5.98\times, and 24.84×24.84\times for FastWAM, ImageWAM and Kairos, respectively. Furthermore, real-world robotic deployments using the FastWAM backbone confirm a consistent 7.44×\times speedup while maintaining close success rate to the multi-step baseline. Code is avaliable at https://github.com/changdong1687/1Step-WAM.