TraceViT: Grounded Trace Supervision for Visual Abstract Reasoning

• URL: http://arxiv.org/abs/2607.29586v1
• Date: Fri, 31 Jul 2026 16:09:44 GMT
• Title: TraceViT: Grounded Trace Supervision for Visual Abstract Reasoning
• Abstract summary: We introduce TraceViT, a looped visual reasoner trained with semantically monotonic transformation chains.<n>TraceViT achieves 67.8% pass@2 on ARC-AGI-1 and 24.3% on ARC-AGI-2.
• Score: 5.703211775506213
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The Abstraction and Reasoning Corpus (ARC) tests whether a model can infer an unseen transformation from a few input-output examples and apply it to a new grid. Looped visual reasoners refine predictions over multiple iterations, but conventional training constrains only the final output, leaving intermediate refinements unconstrained. We propose that these refinements should instead follow the transformation step by step. We introduce TraceViT, a looped visual reasoner trained with semantically monotonic transformation chains. We obtain these chains by rewriting and verifying programmatic task implementations, decomposing each solution into intermediate grid states. Each iteration is grounded by a task reference derived from the few-shot demonstrations and an object workspace representing the current grid state. Because these chains may differ in length from the loop, soft trace alignment enforces only their ordering, letting the model allocate iterations freely. TraceViT achieves 67.8% pass@2 on ARC-AGI-1 and 24.3% on ARC-AGI-2. Controlled ablations on ARC-AGI-1 show that trace supervision becomes beneficial only when paired with grounding. Code and data will be available at https://github.com/LiuBinnan/TraceViT.

Related papers

• Verification-Aware Training for Speculative Decoding [57.84976863792784]
  We introduce Verification-Aware Training (VAT), a plug-in framework that simulates verification at every training step and turns the resulting accept and reject patterns into supervision.<n>VAT modifies only the training objective, so it can be layered on top of existing methods without changing the draft architecture, the target model, or the inference procedure.<n>VAT improves average acceptance length by up to 11.4% and wall-clock speedup by up to 8.7%, with consistent gains across math, code, and chat benchmarks.
  arXiv  Detail & Related papers  (2026-08-31T01:42:10Z)
• Toward Semantically-Seeded, Graph-Propagated Impact Analysis Across Software Artifacts: A Vision [0.0]
  We argue for a training-free and interpretable analyzer that fuses both signals over the same embeddings.<n>A small but complete proof-of-concept on a payment subsystem labelled change scenarios shows the mechanism we care about.
  arXiv  Detail & Related papers  (2026-06-17T09:34:56Z)
• Tool-Call Dependency Structure is Linearly Decodable in LLM Agent Residual Streams [1.0971997884861282]
  LLM agents produce trajectories whose calls form a directed dependency graph.<n>Prior structural probes have targeted static code or chain-of-thought text, not an agent's run-time call graph.<n>A low-capacity edge probe on the residual stream of Qwen3-32B decodes the tool-call dependency graph.
  arXiv  Detail & Related papers  (2026-05-25T00:16:32Z)
• Training LLMs for Multi-Step Tool Orchestration with Constrained Data Synthesis and Graduated Rewards [76.49428173793386]
  LLMs must invoke multiple dependent APIs in the correct order while propagating intermediate outputs.<n>Training models to handle such faces two obstacles: existing environments focus on simple per-turn function calls with simulated data, and binary rewards.<n>We present a framework addressing both challenges. First, we construct a reinforcement learning environment backed by a large-scale cache of real API responses, enabling a data synthesis pipeline that samples valid multi-step orchestration traces.<n>Second, we propose a graduated reward design that decomposes correctness into atomic validity and orchestration.
  arXiv  Detail & Related papers  (2026-03-25T18:31:39Z)
• ExecVerify: White-Box RL with Verifiable Stepwise Rewards for Code Execution Reasoning [9.137158235106941]
  We introduce ExecVerify, which goes beyond text imitation by incorporating verifiable white-box rewards derived from execution traces.<n>Experiments demonstrate that a 7B model trained with ExecVerify achieves performance comparable to 32B models on code reasoning benchmarks.
  arXiv  Detail & Related papers  (2026-03-11T18:49:45Z)
• Is a Pure Transformer Effective for Separated and Online Multi-Object Tracking? [36.5272157173876]
  Multi-Object Tracking (MOT) has demonstrated success in short-term association within the separated tracking-by-detection online paradigm.<n>In this paper, we review the concept of trajectory graphs and propose a novel perspective by representing them as directed acyclic graphs.<n>We introduce a concise Pure Transformer (PuTR) to validate the effectiveness of Transformer in unifying short- and long-term tracking for separated online MOT.
  arXiv  Detail & Related papers  (2024-05-23T02:44:46Z)
• Boot and Switch: Alternating Distillation for Zero-Shot Dense Retrieval [50.47192086219752]
  $texttABEL$ is a simple but effective unsupervised method to enhance passage retrieval in zero-shot settings. By either fine-tuning $texttABEL$ on labelled data or integrating it with existing supervised dense retrievers, we achieve state-of-the-art results.
  arXiv  Detail & Related papers  (2023-11-27T06:22:57Z)
• Single-Stage Visual Relationship Learning using Conditional Queries [60.90880759475021]
  TraCQ is a new formulation for scene graph generation that avoids the multi-task learning problem and the entity pair distribution. We employ a DETR-based encoder-decoder conditional queries to significantly reduce the entity label space as well. Experimental results show that TraCQ not only outperforms existing single-stage scene graph generation methods, it also beats many state-of-the-art two-stage methods on the Visual Genome dataset.
  arXiv  Detail & Related papers  (2023-06-09T06:02:01Z)
• Unified Sequence-to-Sequence Learning for Single- and Multi-Modal Visual Object Tracking [64.28025685503376]
  SeqTrack casts visual tracking as a sequence generation task, forecasting object bounding boxes in an autoregressive manner. SeqTrackv2 integrates a unified interface for auxiliary modalities and a set of task-prompt tokens to specify the task. This sequence learning paradigm not only simplifies the tracking framework, but also showcases superior performance across 14 challenging benchmarks.
  arXiv  Detail & Related papers  (2023-04-27T17:56:29Z)
• Learning a Proposal Classifier for Multiple Object Tracking [36.67900094433032]
  We propose a novel proposal-based learnable framework, which models MOT as a proposal generation, proposal scoring and trajectory inference paradigm on an affinity graph. We experimentally demonstrate that the proposed method achieves a clear performance improvement in both MOTA and IDF1 with respect to previous state-of-the-art on two public benchmarks.
  arXiv  Detail & Related papers  (2021-03-14T10:46:54Z)
• Recurrent Multi-view Alignment Network for Unsupervised Surface Registration [79.72086524370819]
  Learning non-rigid registration in an end-to-end manner is challenging due to the inherent high degrees of freedom and the lack of labeled training data. We propose to represent the non-rigid transformation with a point-wise combination of several rigid transformations. We also introduce a differentiable loss function that measures the 3D shape similarity on the projected multi-view 2D depth images.
  arXiv  Detail & Related papers  (2020-11-24T14:22:42Z)

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.