← Back to all development cases
This case was judged by an anonymized AI reviewer and has not been independently reviewed by people; inspect the draft, cards, and both outputs to assess it.
AutocompleteWith cards better270 / 280 · beu-6f8061abd9548e36

Qwen-3D: A Generalist 3D Vision-Language Model for Spatial Understanding

Computer science · 2608.02980v1

FLOWING EVIDENCE BENCHMARK

How do we compare writing with and without cards?

At the same writing position, the same model produces one first result with cards and one without. All 280 pairs enter the overall anonymized review, which assesses whether a decision not to continue fits the draft.

Overall: 138 with-card wins, 26 ties, 107 no-card wins, 5 pairs where neither was usable, and 4 inconclusive judgments. Both sides wrote text in 218 pairs; 48 had a deliberate stop and 14 included an invalid output.

SAME DRAFT · WITH OR WITHOUT CARDS

01 · FIXED WRITING POSITION

MANUSCRIPT

Same writing position ▌

Matched autocomplete at the same draft position

02 · TWO MATCHED VERSIONS

SHARED BY BOTH

Draft, writing task, and generation model

A · With cards

Relevant literature cards supplied

B · Without cards

No literature cards supplied

Generation model

Same model and version

A → first output

B → first output

Anonymized AI judge

First outputs are anonymized as X and Y

Continuation review: accuracy, fit to the writing task and usability
Returns: X preferred / tie / Y preferred / both unusable

Review again with the answer order swapped

HOW DOES BLIND JUDGING WORK?

① Anonymize both outputs
The AI judge sees the same draft and both first outputs without knowing which received cards.

② Compare and swap order
The judge applies task-specific criteria in X/Y and then Y/X order.

③ Review disagreements
A third pass resolves disagreements. Incomplete reviews remain in the denominator.

How are positions and cards selected?

We selected 40 writing positions in each of seven fields, for 280 comparisons. The position and relevant literature cards were set before each comparison, so both versions began from the same draft.

The main result retains every frozen position, including places where cards might not add a useful next move. This avoids selecting cases based on their outputs, but 49% describes the whole cohort, not a win rate where evidence clearly had an opportunity to help. This cohort did not record that opportunity label for each position before generation, so those two groups cannot yet be separated reliably.

CARD USE · 140/280

Reported card use in continuations with text

The model reports whether a card affected its continuation. In 140 of 280 tests it produced continuation text and reported card use. The previous count of 150 included 10 cases without continuation text and has been corrected.

COMPARISON · 70 / 7 / 30

Which written version was better?

Of the 107 reported-use cases where both versions produced text, the blinded AI judge preferred the with-card version 70 times, found 7 about the same, and preferred the without-card version 30 times.

How are the two table percentages calculated?

ADOPTION COVERAGE

140reported use with text
÷
280all positions
=
50%

WITH-CARD VERSION PREFERRED

70with cards better
÷
107comparable tests
=
65%

The percentages use different groups: 50% counts continuations with text reporting card use among all 280 tests; 65% covers the 107 such cases with text on both sides. The generating model reports card use, while an anonymized AI judge compares writing quality.

Card use is self-reported by the generator. The anonymized AI judge sees the draft and source snippets, so source-aware preference remains possible; there was no independent human review. The overall result describes these 280 frozen positions, not a win probability for any writing task.

Manuscript writing position

Text excerpt · not a PDF page

Research manuscript · excerpt

Qwen-3D: A Generalist 3D Vision-Language Model for Spatial Understanding

Introduction / related work

ttention, and limited context windows prevent long-range spatio- temporal reasoning across frames. Multi-view 3D geom- etry, in the form of depth and camera poses, offers a princi- pled alternative, allowing video frames to be mapped into a shared 3D coordinate system. This enables compression of long multi-view streams into compressed persistent scene representations where temporally distant frames may corre- spond to nearby 3D locations. Existing approaches to integrating such 3D information compression into VLMs in order to improve their long range reasoning abilities generally follow one of two dis- tinct paradigms. One line of work introduces 3D point clouds as auxiliary inputs to the model [11, 19, 20], ei- ther alongside or in place of multi-view images. While these approaches expose explicit geometric structure, they treat point clouds as a modality separate from the visual tokens, overlooking the fact that point cloud features are inherently aligned with image features with corresponding depth. The second line of work integrates geometry di- rectly into the visual token representation. Methods such as LLaV A-3D[58] and Video-3D-LLM[57] modify positional encodings such that

↓ Next paragraph was generated here

The highlighted text is near the writing position. This is a layout of the evaluation input, not a PDF screenshot.

A matched comparison at the same writing positionOverall AI review: With cards better

Both arms use the same model, task and manuscript context. Only the retrieved passages differ. The first recorded output from each arm is shown below.

Read overall AI review rationale

A = without cards · B = with cards

The draft is describing two paradigms for integrating 3D geometry into VLMs. The first paradigm (point clouds as auxiliary inputs) has been fully described. The second paradigm begins with 'Methods such as LLaVA-3D[58] and Video-3D-LLM[57] modify positional encodings such that' — this sentence is incomplete and needs completion. Output B completes this grammatically and conceptually: 'visual tokens are grounded in 3D space rather than 2D image coordinates.' This accurately captures what LLaVA-3D does per the source (it 'utilize[s] the 3D position embeddings to enhance the 2D CLIP Patches with 3D spatial context information'). Output A introduces new problems ('requires retraining the vision encoder and does not generalize across different camera configurations') that are speculative and unsupported by the sources — the LLaVA-3D source mentions no retraining requirement or generalization limitations. Output A also shifts to criticism prematurely, breaking the descriptive flow of the related-work taxonomy. Output B is more accurate, better scoped, and requires no repair.

Evidence

Same model and position · retrieved passages supplied · first output

Preferred by blind review

visual tokens are grounded in 3D space rather than 2D image coordinates.

Passages supplied to the Evidence version

Card 1 · LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D-awareness

Recent advancements in Large Multimodal Models (LMMs) have greatly enhanced their proficiency in 2D visual understanding tasks, enabling them to effectively process and understand images and videos. However, the development of LMMs with 3D scene understanding capabilities has been hindered by the la…

Read full passage excerpt

Recent advancements in Large Multimodal Models (LMMs) have greatly enhanced their proficiency in 2D visual understanding tasks, enabling them to effectively process and understand images and videos. However, the development of LMMs with 3D scene understanding capabilities has been hindered by the lack of large-scale 3D vision-language datasets and powerful 3D encoders. In this paper, we introduce a simple yet effective framework called LLaVA-3D. Leveraging the strong 2D visual understanding priors from LLaVA, our LLaVA-3D efficiently adapts LLaVA for 3D scene understanding without compromising 2D understanding capabilities. To achieve this, we utilize the 3D position embeddings to enhance the 2D CLIP Patches with 3D spatial context information and construct 3D patches. By integrating the 3D position embeddings into 2D LMMs and employing joint 2D and 3D vision-language instruction tuning, we establish a unified architecture for both 2D visual understanding and 3D scene understanding. In …

Card 2 · AffordBot: 3D Fine-grained Embodied Reasoning via Multimodal Large Language Models

Effective human-agent collaboration in physical environments requires understanding not only what to act upon, but also where the actionable elements are and how to interact with them. Existing approaches often operate at the object level or disjointedly handle fine-grained affordance reasoning, lac…

Read full passage excerpt

Effective human-agent collaboration in physical environments requires understanding not only what to act upon, but also where the actionable elements are and how to interact with them. Existing approaches often operate at the object level or disjointedly handle fine-grained affordance reasoning, lacking coherent, instruction-driven grounding and reasoning. In this work, we introduce a new task: Fine-grained 3D Embodied Reasoning, which requires an agent to predict, for each referenced affordance element in a 3D scene, a structured triplet comprising its spatial location, motion type, and motion axis, based on a task instruction. To solve this task, we propose AffordBot, a novel framework that integrates Multimodal Large Language Models (MLLMs) with a tailored chain-of-thought (CoT) reasoning paradigm. To bridge the gap between 3D input and 2D-compatible MLLMs, we render surround-view images of the scene and project 3D element candidates into these views, forming a rich visual represent …

Card 3 · Foundational Models for 3D Point Clouds: A Survey and Outlook

The 3D point cloud representation plays a crucial role in preserving the geometric fidelity of the physical world, enabling more accurate complex 3D environments. While humans naturally comprehend the intricate relationships between objects and variations through a multisensory system, artificial in…

Read full passage excerpt

The 3D point cloud representation plays a crucial role in preserving the geometric fidelity of the physical world, enabling more accurate complex 3D environments. While humans naturally comprehend the intricate relationships between objects and variations through a multisensory system, artificial intelligence (AI) systems have yet to fully replicate this capacity. To bridge this gap, it becomes essential to incorporate multiple modalities. Models that can seamlessly integrate and reason across these modalities are known as foundation models (FMs). The development of FMs for 2D modalities, such as images and text, has seen significant progress, driven by the abundant availability of large-scale datasets. However, the 3D domain has lagged due to the scarcity of labelled data and high computational overheads. In response, recent research has begun to explore the potential of applying FMs to 3D tasks, overcoming these challenges by leveraging existing 2D knowledge. Additionally, language, …

Card 4 · Cog3DMap: Multi-View Vision-Language Reasoning with 3D Cognitive Maps

Precise spatial understanding from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs), as their visual representations are predominantly semantic and lack explicit geometric grounding. While existing approaches augment visual tokens with geometric cues fro…

Read full passage excerpt

Precise spatial understanding from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs), as their visual representations are predominantly semantic and lack explicit geometric grounding. While existing approaches augment visual tokens with geometric cues from visual geometry models, their MLLM is still required to implicitly infer the underlying 3D structure of the scene from these augmented tokens, limiting its spatial reasoning capability. To address this issue, we introduce Cog3DMap, a framework that recurrently constructs an explicit 3D memory from multi-view images, where each token is grounded in 3D space and possesses both semantic and geometric information. By feeding these tokens into the MLLM, our framework enables direct reasoning over a spatially structured 3D map, achieving state-of-the-art performance on various spatial reasoning benchmarks. Code will be made publicly available.

Card 5 · Occ-VLM: Occupancy Grounded Vision Language Model for Indoor Scene Understanding

Recently, vision-language models (VLMs) have made significant progress in 3D scene understanding, driving advances in applications such as embodied intelligence and robotic vision. However, existing approaches typically either rely directly on explicit 3D inputs (e.g., point clouds or RGB-D sequence…

Read full passage excerpt

Recently, vision-language models (VLMs) have made significant progress in 3D scene understanding, driving advances in applications such as embodied intelligence and robotic vision. However, existing approaches typically either rely directly on explicit 3D inputs (e.g., point clouds or RGB-D sequences), or introduce an additional 3D geometry encoder to derive 3D-aware visual tokens from 2D images. Such designs structurally decouple 3D geometric perception from the rich 2D semantics learned via vision-language pre-training, hindering the development of a unified 3D vision-language representation. In this work, we propose Occ-VLM, a novel framework for 3D scene understanding that operates purely on posed RGB images and employs a single 2D vision encoder. Specifically, Occ-VLM reconstructs 3D scene occupancy as an auxiliary geometric prior, which is utilized to spatially associate foreground 2D tokens with 3D space. These tokens are then decoded by a Large Language Model (LLM) for unified …

Model-reported card use

The model did not report that any card materially affected this completion.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

3D spatial relationships are encoded within the token embeddings themselves, but this requires retraining the vision encoder and does not generalize across different camera configurations.

Earlier card-use comparison

The earlier diagnostic compared only cases with reported card use and text on both sides; its scope differs from the overall AI review above.

This case does not have both with-card continuation text and a model-reported use record. It remains among all 280 tests but does not enter the reported-use comparison.

The passages are input material; their presence does not mean an output used them or that they support every claim. Draft and source passages are excerpted; no simulated scores or PDF appear here.