Pith. sign in

REVIEW 5 cited by

Interleaved-Modal Chain-of-Thought

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2411.19488 v2 pith:VII4EY2G submitted 2024-11-29 cs.CV cs.AIcs.LG

Interleaved-Modal Chain-of-Thought

classification cs.CV cs.AIcs.LG
keywords vlmsicotchain-of-thoughtinterleaved-modalimagepromptingreasoningsteps
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Chain-of-Thought (CoT) prompting elicits large language models (LLMs) to produce a series of intermediate reasoning steps before arriving at the final answer. However, when transitioning to vision-language models (VLMs), their text-only rationales struggle to express the fine-grained associations with the original image. In this paper, we propose an image-incorporated multimodal Chain-of-Thought, named \textbf{Interleaved-modal Chain-of-Thought (ICoT)}, which generates sequential reasoning steps consisting of paired visual and textual rationales to infer the final answer. Intuitively, the novel ICoT requires VLMs to enable the generation of fine-grained interleaved-modal content, which is hard for current VLMs to fulfill. Considering that the required visual information is usually part of the input image, we propose \textbf{Attention-driven Selection (ADS)} to realize ICoT over existing VLMs. ADS intelligently inserts regions of the input image to generate the interleaved-modal reasoning steps with ignorable additional latency. ADS relies solely on the attention map of VLMs without the need for parameterization, and therefore it is a plug-and-play strategy that can be generalized to a spectrum of VLMs. We apply ADS to realize ICoT on two popular VLMs of different architectures. Extensive evaluations of three benchmarks have shown that ICoT prompting achieves substantial performance (up to 14\%) and interpretability improvements compared to existing multimodal CoT prompting methods.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Bridging Interleaved Multi-Modal Reasoning as a Unified Decision Process

    cs.AI 2026-07 conditional novelty 7.0

    BRAID jointly optimizes text and image generation in interleaved multi-modal reasoning by casting the full trajectory as a unified MDP with shared advantages and a VLM process reward.

  2. Optical Reasoning: Rethinking Images as an Expressive Reasoning Medium Beyond Text

    cs.AI 2026-06 unverdicted novelty 7.0

    Optical reasoning encodes rationales in images rather than text, matching or exceeding text-based performance on math, science, and multimodal benchmarks while cutting tokens by 28.57% on language tasks and 16% on mul...

  3. DeepLatent: Think with Images via Parallel Latent Visual Reasoning

    cs.CV 2026-05 unverdicted novelty 7.0

    DeepLatent introduces a parallel latent visual reasoning framework with learnable 2D tokens and continuous RL, trained via distillation then RL, plus a new 180K dataset, claiming SOTA benchmark results.

  4. Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space

    cs.CV 2025-12 unverdicted novelty 7.0

    DMLR performs dynamic visual-textual interleaving in latent space using confidence-guided latent policy gradient optimization and a dynamic visual injection strategy, yielding improved multimodal reasoning on benchmarks.

  5. One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding

    cs.CV 2026-04 unverdicted novelty 6.0

    XComp reaches extreme video compression (one token per selective frame) via learnable progressive token compression and question-conditioned frame selection, lifting LVBench accuracy from 42.9 percent to 46.2 percent ...