Detecting Alzheimer’s disease (AD) from narrative transcripts challenges large language models (LLMs): pre-training rarely covers this out-of-distribution task, and all transcript demos describe the same scene, producing highly homogeneous contexts. These factors cripple both the model’s built-in task knowledge (task cognition) and its ability to surface subtle, class-discriminative cues (contextual perception). Because cognition is fixed after pre-training, improving in-context learning (ICL) for AD detection hinges on enriching perception through better demonstration (demo) sets. We demonstrate that standard ICL quickly saturates, its demos lack diversity (context width) and fail to convey fine-grained signals (context depth), and that recent task vector (TV) approaches improve broad task adaptation by injecting TV into the LLMs' hidden states (HSs), they are ill-suited for AD detection due to the mismatch of injection granularity, strength and position. To address these bottlenecks, we introduce DA4ICL, a demo-centric anchoring framework that jointly expands context width via Diverse and Contrastive Retrieval (DCR) and deepens each demo's signal via Projected Vector Anchoring (PVA) at every Transformer layer. Across three AD benchmarks, DA4ICL achieves large, stable gains over both ICL and TV baselines, charting a new paradigm for fine-grained, OOD and low-resource LLM adaptation.

Visual Question Answering (VQA) has increasingly shifted from direct prediction to step-by-step reasoning, driven by advances in Multi-modal Large Language Models (MLLMs) and Multi-modal Chain-of-Thought (MCoT). However, small MLLMs (S-MLLMs) often suffer from a "Faithfulness Gap", where they arrive at correct answers with inconsistent or hallucinatory intermediate reasoning processes. In this work, we first conduct a systematic diagnostic of MCoT consistency and identify two prevalent patterns that undermine conventional trajectory-level optimization: Intermediate Error and Inconsistent Reasoning. We categorize intermediate errors into Caption Errors—Hallucination, Contradiction, and Irrelevance, and Reasoning Errors—Divergence and Spuriousness. Driven by these findings, we curate MCoT-SP, a Stage-wise Preference dataset with severity-aware annotations constructed through an automated pipeline that detects and corrects intermediate errors using entailment-based checks and LLM-based correction, followed by post-verification to ensure the reliability of the preference pairs. Leveraging this resource, we propose Intermediate Process Direct Preference Optimization (IP-DPO), which performs stage-wise preference optimization by treating each intermediate stage as independent preference units, while jointly integrating preference learning with supervised fine-tuning on preferred complete MCoT responses to preserve global coherence. Extensive evaluations on LLaVA-Next and Qwen2.5-VL demonstrate that IP-DPO outperforms standard DPO and SFT baselines across two benchmarks spanning six VQA datasets. Our analysis indicates that improvements in intermediate consistency substantially exceed the corresponding improvements in final answer accuracy, revealing that IP-DPO primarily improves the underlying reasoning process rather than merely overfitting to target labels.

Detecting Alzheimer's Disease (AD) from narrative transcripts remains a challenging task for large language models (LLMs), particularly under out-of-distribution (OOD) and datascarce conditions. While in-context learning (ICL) provides a parameter-efficient alternative to fine-tuning, existing ICL approaches often suffer from task recognition failure, suboptimal demonstration selection, and misalignment between label words and task objectives, issues that are amplified in clinical domains like AD detection. We propose Explicit Knowledge In-Context Learners (EK-ICL), a novel framework that integrates structured explicit knowledge to enhance reasoning stability and task alignment in ICL. EK-ICL incorporates three knowledge components: confidence scores derived from small language models (SLMs) to ground predictions in task-relevant patterns, parsing feature scores to capture structural differences and improve demo selection, and label word replacement to resolve semantic misalignment with LLM priors. In addition, EK-ICL employs a parsing-based retrieval strategy and ensemble prediction to mitigate the effects of semantic homogeneity in AD transcripts. Extensive experiments across three AD datasets demonstrate that EK-ICL significantly outperforms state-of-the-art fine-tuning and ICL baselines. Further analysis reveals that ICL performance in AD detection is highly sensitive to the alignment of label semantics and task-specific context, underscoring the importance of explicit knowledge in clinical reasoning under low-resource conditions.

Pre-trained Large Language Models (LLMs) have demonstrated prominent generalization to various linguistic tasks. However, due to the inherent modality and task discrepancy, parameter-efficient transfer learning for adapting LLMs to vision-language (VL) tasks remains challenging, which may struggle with excessive extra computation and data expenditure for VL pre-training and disconnection between multi-modal representations. This paper concentrates on the parameter-efficient adaptation of LLMs to VL tasks without inflexible multi-modal alignment pre-training on additional image-text pairs. Inspired by Instruction Tuning and the nature of multi-modal representation learning, we propose Multi-modal Prompt Tuning for Language Models (MPT4LM). This method provides text-relevant visual prompts via a plug-and-play Cross-Attention module and integrates them with textual Learnable Instruction as multi-modal prompts into LLMs. We further assemble MPT4LM with the currently prevalent Adapter approach to alleviate the trainable parameter scale and facilitate the collaboration of multi-modal prompts. We evaluate MPT4LM upon two representative LLMs: LLAMA-2 and Flan-T5, over two VL tasks: Visual Question Answering (VQAv2.0, GQA) and Visual Entailment (SNLI-VE). Extensive experimental results reveal that MPT4LM achieves state-of-the-art performance among prompting methods with only fine-tuning about 0.65% of the parameters of backbones, indicating a better trade-off between computation and data overhead and model performance. Our code is available at: https://github.com/YzM1a0/MPT4LM.

Prompt tuning, like CoOp, has recently shown promising vision recognizing and transfer learning ability on various downstream tasks with the emergence of large pre-trained vision-language models like CLIP. However, we identify that existing uni-modal prompt tuning approaches may result in sub-optimal performance since this uni-modal design breaks the original alignment of textual and visual representations in the pre-trained model. Inspired by the nature of pre-trained vision-language models, we aim to achieve completeness in prompt tuning and propose a novel approach called Multi-modal Deep-symphysis Prompt Tuning, dubbed as MuDPT, which extends independent multi-modal prompt tuning by additionally learning a model-agnostic transformative network to allow deep hierarchical bi-directional prompt fusion. We evaluate the effectiveness of MuDPT on few-shot vision recognition and out-of-domain generalization tasks. Compared with the state-of-the-art methods, MuDPT achieves better recognition and generalization ability with an apparent margin thanks to synergistic alignment of textual and visual representations. Our code is available at: https://github.com/Mechrev0/MuDPT.