Neural Information Processing Systems http://nips.cc/

With the continued advancement of Large Language Models (LLMs) Agents in reasoning, planning, and decision-making, benchmarks have become crucial in evaluating these skills. However, there is a notable gap in benchmarks for real-time strategic decision-making. StarCraft II (SC2), with its complex and dynamic nature, serves as an ideal setting for such evaluations. To this end, we have developed TextStarCraft II, a specialized environment for assessing LLMs in real-time strategic scenarios within SC2. Addressing the limitations of traditional Chain of Thought (CoT) methods, we introduce the Chain of Summarization (CoS) method, enhancing LLMs' capabilities in rapid and effective decision-making. Commercial Model Knowledge: Evaluated four commercial models on SC2 knowledge; GPT-4 ranked highest by Grandmaster-level experts.3.

Addressing the challenge of effectively processing long contexts has become a critical issue for Large Language Models (LLMs). Two common strategies have emerged: 1) reducing the input length, such as retrieving relevant chunks by Retrieval-Augmented Generation (RAG), and 2) expanding the context window limit of LLMs. However, both strategies have drawbacks: input reduction has no guarantee of covering the part with needed information, while window extension struggles with focusing on the pertinent information for solving the task. To mitigate these limitations, we propose Chain-of-Agents (CoA), a novel framework that harnesses multi-agent collaboration through natural language to enable information aggregation and context reasoning across various LLMs over long-context tasks. CoA consists of multiple worker agents who sequentially communicate to handle different segmented portions of the text, followed by a manager agent who synthesizes these contributions into a coherent final output. CoA processes the entire input by interleaving reading and reasoning, and it mitigates long context focus issues by assigning each agent a short context.

Diffusion models have significantly advanced the field of generative modeling. However, training a diffusion model is computationally expensive, creating a pressing need to adapt off-the-shelf diffusion models for downstream generation tasks. Current fine-tuning methods focus on parameter-efficient transfer learning but overlook the fundamental transfer characteristics of diffusion models. In this paper, we investigate the transferability of diffusion models and observe a monotonous chain of forgetting trend of transferability along the reverse process. Based on this observation and novel theoretical insights, we present Diff-Tuning, a frustratingly simple transfer approach that leverages the chain of forgetting tendency.

Large Language Models (LLMs) have demonstrated remarkable performance in solving complex reasoning tasks through mechanisms like Chain-of-Thought (CoT) prompting, which emphasizes verbose, step-by-step reasoning. However, humans typically employ a more efficient strategy: drafting concise intermediate thoughts that capture only essential information. In this work, we propose Chain of Draft (CoD), a novel paradigm inspired by human cognitive processes, where LLMs generate minimalistic yet informative intermediate reasoning outputs while solving tasks. By reducing verbosity and focusing on critical insights, CoD matches or surpasses CoT in accuracy while using as little as only 7.6% of the tokens, significantly reducing cost and latency across various reasoning tasks. Our code and data are available at https://github.com/sileix/chain-of-draft.

The use of Large Language Models (LLMs) for code generation has gained significant attention in recent years. Existing methods often aim to improve the quality of generated code by incorporating additional contextual information or guidance into input prompts. Many of these approaches adopt sequential reasoning strategies, mimicking human-like step-by-step thinking. However, such strategies may constrain flexibility, as they do not always align with the structured characteristics of programming languages. This paper introduces the Chain of Grounded Objectives (CGO), a method that embeds functional objectives into input prompts to enhance code generation. By leveraging appropriately structured objectives as input and avoiding explicit sequential procedures, CGO adapts effectively to the structured nature of programming tasks. Empirical evaluations demonstrate that CGO effectively enhances code generation, addressing limitations of existing approaches.

Paper Summary: This paper presented a novel approach that performs chain of reasonings on the object level to generate answer for visual question answering. Object-level visual embeddings are first extracted through object detection networks as visual representation and sentence embedding of the question are extract question representation. Based on these, a sequential model that performs multi-steps of relational inference over (compound) object embeddings with the guidance of question is used to obtain the final representation for each sub-chain inference. A concatenation of these embeddings are then used to perform answer classification. Extensive experiments have been conducted on four public datasets and it achieves state-of-the-art performance on all of them.

Are you sure you want to exit? Any unsave changes will be lost.

So far, however, Domino's has stopped short of the latest McDonald's play: acquiring entire tech start-ups. In March, McDonald's spent more than $300 million to buy Dynamic Yield, the Tel Aviv-based company that developed the artificial intelligence tools now used at thousands of McDonald's drive-throughs. The deal "has changed the way the high-tech industry thinks about potential M&A," said Liad Agmon, a former Israeli intelligence official who co-founded Dynamic Yield. "We'll see more nontraditional tech companies buying tech companies as an accelerator for their digital efforts. It was genius on McDonald's side."

The difficulty of finding information on the World Wide Web by browsing hypertext documents has led to the development and deployment of various search engines and indexing techniques. However, many information-gathering tasks are better handled by finding a referral to a human expert rather than by simply interacting with online information sources. A personal referral allows a user to judge the quality of the information he or she is receiving as well as to potentially obtain information that is deliberately not made public. The process of finding an expert who is both reliable and likely to respond to the user can be viewed as a search through the network of social relationships between individuals as opposed to a search through the network of hypertext documents. Project is to create models of social networks by data mining the web and develop tools that use the models to assist in locating experts and related information search and evaluation tasks.