Large Language Models (LLMs) require continuous updates to maintain accurate and current knowledge as the world evolves. While existing knowledge editing approaches offer various solutions for knowledge updating, they often struggle with sequential editing scenarios and harm the general capabilities of the model, thereby significantly hampering their practical applicability.This paper proposes a two-stage framework combining robust supervised fine-tuning (R-SFT) with model merging for knowledge editing. Our method first fine-tunes the LLM to internalize new knowledge fully, then merges the fine-tuned model with the original foundation model to preserve newly acquired knowledge and general capabilities. Experimental results demonstrate that our approach significantly outperforms existing methods in sequential editing while better preserving the original performance of the model, all without requiring any architectural changes. Code is available at [Applied-Machine-Learning-Lab/MM4KE](https://github.com/Applied-Machine-Learning-Lab/MM4KE).

Large Language Models (LLMs) have made substantial strides in a broad array of natural language tasks. Recently, LLMs have demonstrated potential reasoning capabilities through prompt design, such as the Chain of Thought (CoT). Despite their superiority in question answering, LLMs still face challenges in answering questions that require multi-hop reasoning, often generating unreliable reasoning chains during answer generation. To improve LLMs’ performance in multi-hop reasoning, we introduce a novel reasoning approach, AnchorCoT, designed to assist LLMs in answering questions involving complex logical reasoning steps. AnchorCoT first predicts key entities which work as important “anchors” to guide the reasoning process and then employs a novel ranking algorithm to ensure the logical sequence of the predicted answers.We implement AnchorCoT on Qwen2.5-7B/14B and GPT-4o and evaluate our method on widely used multi-hop reasoning datasets, including HotpotQA, 2WikiMultiHopQA, and MuSiQue-Ans. The experimental results show that AnchorCoT outperforms existing methods in multi-hop question reasoning and provides more accurate reasoning results in multi-hop question answering tasks.

Despite the great success of Large Vision-Language Models (LVLMs), they inevitably suffer from hallucination. As we know, both the visual encoder and the Large Language Model (LLM) decoder in LVLMs are Transformer-based, allowing the model to extract visual information and generate text outputs via attention mechanisms. We find that the attention distribution of LLM decoder on image tokens is highly consistent with the visual encoder and both distributions tend to focus on particular background tokens rather than the referred objects in the image. We attribute to the unexpected attention distribution to an inherent flaw in the visual encoder itself, which misguides LLMs to over emphasize the redundant information and generate object hallucination. To address the issue, we propose DAMRO, a novel training-free strategy that **D**ive into **A**ttention **M**echanism of LVLM to **R**educe **O**bject Hallucination. Specifically, our approach employs classification token (CLS) of ViT to filter out high-attention tokens scattered in the background and then eliminate their influence during decoding stage. We evaluate our method on LVLMs including LLaVA-1.5, LLaVA-NeXT and InstructBLIP, using various benchmarks such as POPE, CHAIR, MME and GPT-4V Aided Evaluation. The results demonstrate that our approach significantly reduces the impact of these outlier tokens, thus effectively alleviating the hallucination of LVLMs.