Large language models (LLMs) have shown strong reasoning abilities in mathematical reasoning, commonsense reasoning, and code generation. However, negotiation requires interactive reasoning, where each action can influence the opponent’s future behavior. Agents must therefore look beyond the current dialogue state and consider the long-term effects of their actions. Existing LLM-based negotiation agents often rely on prompting or fine-tuning from self-play trajectories. These methods can produce fluent responses and help models learn useful bargaining patterns. Yet at decision time, they often remain reactive, responding mainly to the current dialogue without reasoning about how different actions may shape future negotiation paths. We introduce an online strategic reasoning method that combines Monte Carlo Tree Search (MCTS)-based lookahead planning with opponent simulation. Before responding, our agent uses a proxy opponent agent to simulate future interactions and evaluate candidate strategies by their expected long-term outcomes. To make opponent simulation more precise, we proposed a Theory-of-Mind-inspired opponent modeling module that infers latent opponent attributes from dialogue history and conditions the proxy opponent agent during search. Experiments on CraigslistBargain and A2A-NT show that our method improves negotiation efficiency and profitability.

Retrieval-augmented generation (RAG) enables large language models (LLMs) to dynamically access external information, which is powerful for answering questions over previously unseen documents. Nonetheless, they struggle with high-level conceptual understanding and holistic comprehension due to limited context windows, which constrain their ability to perform deep reasoning over long-form, domain-specific content such as full-length books. To solve this problem, knowledge graphs (KGs) have been leveraged to provide entity-centric structure and hierarchical summaries, offering more structured support for reasoning. However, existing KG-based RAG solutions remain restricted to text-only inputs and fail to leverage the complementary insights provided by other modalities such as vision. On the other hand, reasoning from visual documents requires textual, visual, and spatial cues into structured, hierarchical concepts. To address this issue, we introduce a multimodal knowledge graph-based RAG that enables cross-modal reasoning for better content understanding. Our method incorporates visual cues into the construction of knowledge graphs, the retrieval phase, and the answer generation process. Experimental results across both global and fine-grained question answering tasks show that our approach consistently outperforms existing approaches on both textual and multimodal benchmarks.

Prompt Tuning has been a popular Parameter-Efficient Fine-Tuning method attributed to its remarkable performance with few updated parameters on various large-scale pretrained Language Models (PLMs). Traditionally, each prompt has been considered indivisible and updated independently, leading the parameters increase proportionally as prompt length grows. To address this issue, we propose Adaptive Codebook for Composite and Efficient Prompt Tuning (ACCEPT). In our method, we refer to the concept of product quantization (PQ), allowing all soft prompts to share a set of learnable codebook vectors in each subspace, with each prompt differentiated by a set of adaptive weights. We achieve the superior performance on 17 diverse natural language tasks including natural language understanding (NLU) and question answering (QA) tasks by tuning only 0.3% of parameters of the PLMs. Our approach also excels in few-shot and large model settings, highlighting its significant potential.