LLM-as-a-Judge offers a promising alternative to human judges across various tasks, yet inherent biases, particularly position bias - a systematic preference for answers based on their position in the prompt - compromise its effectiveness. Our study investigates this issue by developing a framework to systematically study and quantify position bias using metrics such as repetitional consistency, positional consistency, and positional fairness. We conduct experiments with 9 judge models across 22 tasks from the MTBench and DevBench benchmarks and nearly 40 answer-generating models, generating approximately 80,000 evaluation instances. This comprehensive assessment reveals significant variations in bias across judges and tasks. Although GPT-4 often excels in positional consistency and fairness, some more cost-effective models perform comparably or even better in specific tasks, highlighting essential trade-offs between consistency, fairness, and cost. Our results also demonstrate high consistency of judgment across repetitions, confirming that position bias is not due to random variations. This research significantly contributes to the field by introducing new concepts for understanding position bias and providing a multi-dimensional framework for evaluation. These insights guide the selection of optimal judge models, enhance benchmark design, and lay the foundation for future research into effective debiasing strategies, ultimately enhancing the reliability of LLM evaluators.

Grounded Multimodal Named Entity Recognition (GMNER) task aims to identify named entities, entity types and their corresponding visual regions. GMNER task exhibits two challenging attributes: 1) The tenuous correlation between images and text on social media contributes to a notable proportion of named entities being ungroundable. 2) There exists a distinction between coarse-grained noun phrases used in similar tasks (e.g., phrase localization) and fine-grained named entities. In this paper, we propose RiVEG, a unified framework that reformulates GMNER into a joint MNER-VE-VG task by leveraging large language models (LLMs) as connecting bridges. This reformulation brings two benefits: 1) It enables us to optimize the MNER module for optimal MNER performance and eliminates the need to pre-extract region features using object detection methods, thus naturally addressing the two major limitations of existing GMNER methods. 2) The introduction of Entity Expansion Expression module and Visual Entailment (VE) module unifies Visual Grounding (VG) and Entity Grounding (EG). This endows the proposed framework with unlimited data and model scalability. Furthermore, to address the potential ambiguity stemming from the coarse-grained bounding box output in GMNER, we further construct the new Segmented Multimodal Named Entity Recognition (SMNER) task and corresponding Twitter-SMNER dataset aimed at generating fine-grained segmentation masks, and experimentally demonstrate the feasibility and effectiveness of using box prompt-based Segment Anything Model (SAM) to empower any GMNER model with the ability to accomplish the SMNER task. Extensive experiments demonstrate that RiVEG significantly outperforms SoTA methods on four datasets across the MNER, GMNER, and SMNER tasks.

In this paper, we present the LingOly benchmark, a novel benchmark for advanced reasoning abilities in large language models. Using challenging Linguistic Olympiad puzzles, we evaluate (i) capabilities for in-context identification and generalisation of linguistic patterns in very low-resource or extinct languages, and (ii) abilities to follow complex task instructions. The LingOly benchmark covers more than 90 mostly low-resource languages, minimising issues of data contamination, and contains 1,133 problems across 6 formats and 5 levels of human difficulty. We assess performance with both direct accuracy and comparison to a no-context baseline to penalise memorisation. Scores from 11 state-of-the-art LLMs demonstrate the benchmark to be challenging, and models perform poorly on the higher difficulty problems. On harder problems, even the top model only achieved 38.7% accuracy, 24.7% improvement over the no-context baseline. Large closed models typically outperform open models, and in general, the higher resource the language, the better the scores. These results indicate, in absence of memorisation, true multi-step out-of-domain reasoning remains a challenge for current language models.

A representation learning method is considered stable if it consistently generates similar representation of the given data across multiple runs. Word Embedding Methods (WEMs) are a class of representation learning methods that generate dense vector representation for each word in the given text data. The central idea of this paper is to explore the stability measurement of WEMs using intrinsic evaluation based on word similarity. We experiment with three popular WEMs: Word2Vec, GloVe, and fastText. For stability measurement, we investigate the effect of five parameters involved in training these models. We perform experiments using four real-world datasets from different domains: Wikipedia, News, Song lyrics, and European parliament proceedings. We also observe the effect of WEM stability on three downstream tasks: Clustering, POS tagging, and Fairness evaluation. Our experiments indicate that amongst the three WEMs, fastText is the most stable, followed by GloVe and Word2Vec.

This work is about estimating the hallucination rate for in-context learning (ICL) with Generative AI. In ICL, a conditional generative model (CGM) is prompted with a dataset and asked to make a prediction based on that dataset. The Bayesian interpretation of ICL assumes that the CGM is calculating a posterior predictive distribution over an unknown Bayesian model of a latent parameter and data. With this perspective, we define a \textit{hallucination} as a generated prediction that has low-probability under the true latent parameter. We develop a new method that takes an ICL problem -- that is, a CGM, a dataset, and a prediction question -- and estimates the probability that a CGM will generate a hallucination. Our method only requires generating queries and responses from the model and evaluating its response log probability. We empirically evaluate our method on synthetic regression and natural language ICL tasks using large language models.

The capacity of Large Language Models (LLMs) to comprehend and reason over long contexts is pivotal for advancements in diverse fields. Yet, they still stuggle with capturing long-distance dependencies within sequences to deeply understand semantics. To address this issue, we introduce Query-aware Inference for LLMs (Q-LLM), a system designed to process extensive sequences akin to human cognition. By focusing on memory data relevant to a given query, Q-LLM can accurately capture pertinent information within a fixed window size and provide precise answers to queries. It doesn't require extra training and can be seamlessly integrated with any LLMs. Q-LLM using LLaMA3 (QuickLLaMA) can read Harry Potter within 30s and accurately answer the questions. Q-LLM improved by 7.17% compared to the current state-of-the-art on LLaMA3, and by 3.26% on Mistral on the $\infty$-bench. In the Needle-in-a-Haystack task, On widely recognized benchmarks, Q-LLM improved upon the current SOTA by 7.0% on Mistral and achieves 100% on LLaMA3. Our code can be found in https://github.com/dvlab-research/Q-LLM.

The beloved elephant Osh has celebrated his 30th birthday with an assortment of delightful treats, including watermelons, popsicles, peanut butter and bran snow cones and even a personalized piñata. A recently-published study claims that the sounds of African elephants may have a lot more significance than humans think. The research, which was published in a journal called Nature Ecology and Evolution on Monday, found that African elephants call each other unique names. The study explains that researchers followed elephants around to observe how they communicated to each other, particularly by taking careful note of which elephants called out sounds and which elephants appeared to respond. The names came in the form of low rumbles, which elephants can hear from long distances.

Catch up on your favourite BBC radio show from your favourite DJ right here, whenever you like. Listen without limits with BBC Sounds.

If you've been alive between Christmas and New Years, you've probably read a Best of the Year list. Last year, according to The New York Times, Víkingur Ólafsson's recording of Bach's "Goldberg Variations," the actor Bella Ramsey, and a sushi-and-scuba video game called Dave the Diver were worthy of your time and attention. These annual rundowns arrive during a period of reflection, when a full year's worth of human art and industry is about to recede into history. A new take on this list has now emerged: the Best So Far list. What does it mean to offer an account of the best culture of … the first five months of 2024?

Renée DiResta is the research manager of the Stanford Internet Observatory and the author of Invisible Rulers: The People Who Turn Lies into Reality. At the end of May, OpenAI marked a new "first" in its corporate history. It wasn't an even more powerful language model or a new data partnership, but a report disclosing that bad actors had misused their products to run influence operations. The company had caught five networks of covert propagandists--including players from Russia, China, Iran, and Israel--using their generative AI tools for deceptive tactics that ranged from creating large volumes of social media comments in multiple languages to turning news articles into Facebook posts. The use of these tools, OpenAI noted, seemed intended to improve the quality and quantity of output.