China

Recent work on reasoning‐oriented language models, exemplified by \textit{o1-like} systems, suggests that reinforcement-learning (RL) finetuning does not create new capabilities but instead strengthens reasoning patterns already latent in the pretrained network. We test this claim by training \emph{steering vectors}: layer-wise biases that additively amplify selected hidden features while leaving all original weights unchanged. Experiments on four base models across the GSM8K and MATH benchmarks show that steering vectors recover, and in several cases exceed, the accuracy of fully-tuned counterparts. This result supports the view that the required reasoning skills pre-exist in the base model. Further, logit-lens analysis reveals that the trained vectors consistently boost token groups linked to structured languages and logical connectors, providing an interpretable account that aligns with the demands of quantitative reasoning tasks.

EMNLP 2025

Steering LLM Reasoning Through Bias-Only Adaptation

parameter-efficient-training

knowledge tracing/discovering/inducing

chain-of-thought

reasoning

# Welcome everyone to ACL 2022!

The 60th Annual Meeting of the Association for Computational Linguistics is taking place May 22-27, 2022 as a hybrid event, in Dublin and online. We are happy to welcome all of you to this anniversary edition with an almost 50-50 in-person and virtual participation. 
The main conference program features oral presentations, in-person and virtual posters and demo sessions, a plenary session for our best paper presentations and awards, three amazing keynote events and two new initiatives of invited talks: Spotlight Talks for Young Rising Stars (STIRS) and The Next Big Idea Talks. Posters (including Findings of ACL 2022) and demos are grouped by areas for both the in-person and the virtual sessions. For the virtual component, the talks will be on Zoom and the posters and the demos will be in GatherTown. The Student Research Workshop will have an oral session and a poster session as part of Poster Session 1. The program also features eight Tutorials and 28 Workshops. 

 
We wish you a wonderful conference! 
[**The ACL 2022 Organizing Committee**](https://www.2022.aclweb.org/organisers)
 
[**Conference Handbook**](https://drive.google.com/file/d/1_BUCMfhMVrjG9E2e71aHdHeE28KSje0l/view?usp=sharing) 
[**Mini Handbook**](https://drive.google.com/file/d/1qlBKl0wzmlVF1oCeMQl3BahLd9nLP5Ce/view?usp=sharing) 
[**Posters and Demo guides**](https://drive.google.com/file/d/1UucMAoCNncIOaH1rMMDa0owuG9qgvJTG/view?usp=sharing)

ACL 2022

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing (NLP). 

poster

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

Human feedback is crucial in the interactions between humans and Large Language Models (LLMs). However, existing research primarily focuses on benchmarking LLMs in single-turn dialogues. Even in benchmarks designed for multi-turn dialogues, the user utterances are often independent, neglecting the nuanced and complex nature of human feedback within real-world usage scenarios. To fill this research gap, we introduce FB-Bench, a fine-grained, multi-task benchmark designed to evaluate LLMs' responsiveness to human feedback under real-world usage scenarios in Chinese. Drawing from the two main interaction scenarios, FB-Bench comprises 591 meticulously curated samples, encompassing eight task types, five deficiency types of response, and nine feedback types. We extensively evaluate a broad array of popular LLMs, revealing significant variations in their performance across different interaction scenarios. Further analysis indicates that task, human feedback, and deficiencies of previous responses can also significantly impact LLMs' responsiveness. Our findings underscore both the strengths and limitations of current models, providing valuable insights and directions for future research.

FB-Bench: A Fine-Grained Multi-Task Benchmark for Evaluating LLMs' Responsiveness to Human Feedback

As large language models (LLMs) increasingly exhibit human-like capabilities, a fundamental question emerges: How can we enable LLMs to learn the underlying patterns from limited examples in entirely novel environments and apply them effectively? This question is central to the ability of LLMs in inductive reasoning. Existing research on LLM-based inductive reasoning can be broadly categorized based on whether the underlying rules are expressible via explicit mathematical equations. However, many recent studies in the beyond-equations category have emphasized rule design without grounding them in specific scenarios. Inspired by the parallels between inductive reasoning and human scientific discovery, we propose the task of LLM-Based Scientific Inductive Reasoning Beyond Equations and introduce a new benchmark, SIRBench-V1, to evaluate the inductive reasoning abilities of LLMs in scientific settings. Our experimental results show that current LLMs still struggle with this task, underscoring its difficulty and the need for further advancement in this area.

On LLM-Based Scientific Inductive Reasoning Beyond Equations

As interest grows in generating long, detailed image captions, existing automatic evaluation metrics are increasingly strained. N-gram-based metrics though efficient, fail to capture semantic correctness, especially for longer outputs. Representational Similarity (RS) metrics, designed to address this, initially saw limited use due to high computational costs, while today, despite advances in hardware, they remain unpopular as they fall short even of weak baselines such as BLEU. Meanwhile, metrics based on large language models (LLMs) show strong correlation with human judgments, but remain too expensive for use in model development. We introduce SPECS (Specificity-Enhanced CLIP-Score), a reference-free RS metric tailored for long image captioning. SPECS modifies CLIP with a new objective that emphasizes specificity: rewarding correct details and penalizing errors. We show that SPECS matches the performance of leading LLM-based metrics in correlating with human judgments, while being far more efficient. This makes it a practical alternative for iterative checkpoint evaluation during image captioning model development.

SPECS: Specificity-Enhanced CLIP-Score for Long Image Caption Evaluation

Recent progress in Large Language Models (LLMs) has opened new avenues for solving complex optimization problems, including Neural Architecture Search (NAS). However, existing LLM-driven NAS approaches rely heavily on prompt engineering and domain-specific tuning, limiting their practicality and scalability across diverse tasks. In this work, we propose LM-Searcher, a novel framework that leverages LLMs for cross-domain neural architecture optimization without the need for extensive domain-specific adaptation. Central to our approach is NCode, a universal numerical string representation for neural architectures, which enables cross-domain architecture encoding and search. We also reformulate the NAS problem as a ranking task, training LLMs to select high-performing architectures from candidate pools using instruction-tuning samples derived from a novel pruning-based subspace sampling strategy. Our curated dataset, encompassing a wide range of architecture-performance pairs, encourages robust and transferable learning. Comprehensive experiments demonstrate that LM-Searcher achieves competitive performance in both in-domain (e.g., CNNs for image classification) and out-of-domain (e.g., LoRA configurations for segmentation and generation) tasks, establishing a new paradigm for flexible and generalizable LLM-based architecture search.

LM-Searcher: Cross-domain Neural Architecture Search with LLMs via Unified Numerical Encoding

Large language models (LLMs) now support context windows exceeding 128K tokens, but this comes with significant memory requirements and high inference latency. Quantization can mitigate these costs, but may degrade performance. We present the first large-scale evaluation of quantized LLMs on tasks with long-inputs (>64K tokens) and long-form outputs. Our evaluation spans 9.7K examples, five quantization methods (FP8, GPTQ-int8, AWQ-int4, GPTQ-int4, BNB-nf4), and five models (Llama-3.1 8B and 70B; Qwen-2.5 7B, 32B, and 72B). Results show that, on average, 8-bit quantization preserves accuracy (<0.8% drop), whereas 4-bit methods incur substantial losses up to 59% on long context tasks. Performance degradation from quantization is more pronounced in long-input tasks than in long-form generations. These drops are further amplified in a multilingual setup. Furthermore, the impact of quantization varies across models. While Qwen-2.5 72B remains robust under BNB-nf4, Llama-3.1 70B suffers a 32% performance drop. These findings underscore the importance of rigorous evaluation before deploying quantized LLMs, especially in long‑context and multilingual settings.

Does quantization affect models’ performance on long-context tasks?

*The comparison between discriminative and generative classifiers has intrigued researchers since [Efron (1975)’s](https://www.jstor.org/stable/2285453) seminal analysis of logistic regression versus discriminant analysis. While early theoretical work established that generative classifiers exhibit lower sample complexity but higher asymptotic error in simple linear settings, these trade-offs remain unexplored in the transformer era. We present the first comprehensive evaluation of modern generative and discriminative architectures—Auto-regressive, Masked Language Modeling, Discrete Diffusion, and Encoders for text classification. Our study reveals that the classical ``two regimes" phenomenon manifests distinctly across different architectures and training paradigms. Beyond accuracy, we analyze sample efficiency, calibration, noise robustness, and ordinality across diverse scenarios. Our findings offer practical guidance for selecting the most suitable modeling approach based on real-world constraints such as latency and data limitations.*

Generative or Discriminative? Revisiting Text Classification in the Era of Transformers

Recent advances in large language models (LLMs) have demonstrated strong mathematical reasoning abilities, even in visual contexts, with some models surpassing human performance on existing benchmarks. However, these benchmarks lack structured age categorization, clearly defined skill requirements, and—crucially—were not designed to assess human performance in international competitions. To address these limitations, we introduce MathGames, a new benchmark of 2,183 high-quality mathematical problems (both text-only and multimodal) in an open-ended format, sourced from an international mathematical games championships. Spanning seven age groups and a skill-based taxonomy, MathGames enables a structured evaluation of LLMs' mathematical and logical reasoning abilities. Our experiments reveal a substantial gap between state-of-the-art LLMs and human participants—even 11-year-olds consistently outperform some of the strongest models—highlighting the need for advancements. Further, our detailed error analysis offers valuable insights to guide future research. The data is publicly available at https://anonymous.4open.science/r/math-games/.

Can Large Language Models Win the International Mathematical Games?

We introduce UniversalCEFR, a large-scale multilingual multidimensional dataset of texts annotated according to the CEFR (Common European Framework of Reference) scale in 13 languages. To enable open research in both automated readability and language proficiency assessment, UniversalCEFR comprises 505,807 CEFR-labeled texts curated from educational and learner-oriented resources, standardized into a unified data format to support consistent processing, analysis, and modeling across tasks and languages. To demonstrate its utility, we conduct benchmark experiments using three modelling paradigms: a) linguistic feature-based classification, b) fine-tuning pre-trained LLMs, and c) descriptor-based prompting of instruction-tuned LLMs. Our results further support using linguistic features and fine-tuning pretrained models in multilingual CEFR level assessment. Overall, UniversalCEFR aims to establish best practices in data distribution in language proficiency research by standardising dataset formats and promoting their accessibility to the global research community.

UniversalCEFR: Enabling Open Multilingual Research on Language Proficiency Assessment

Recent advances in vision-language models (VLMs) have enabled accurate image-based geolocation, raising serious concerns about location privacy risks in everyday social media posts. However, current benchmarks remain coarse-grained, linguistically biased, and lack multimodal and privacy-aware evaluations. To address these gaps, we present KoreaGEO Bench, the first benchmark designed for fine-grained, multimodal, and privacy-aware evaluation of VLM geolocation, using Korean street views as a rich case study. Our benchmark dataset comprises 1,080 high-resolution images sampled across four socio-spatial clusters and nine place types, enriched with multi-contextual annotations and two styles of Korean captions simulating real-world privacy exposure. We introduce a three-path evaluation protocol to assess ten mainstream VLMs under varying input modalities and analyze their accuracy, spatial bias, and reasoning behavior. Results reveal modality-driven shifts in localization precision and highlight structural prediction biases toward core cities. Ultimately, our work calls for a dual approach in geolocation benchmarking: alongside pursuing the breadth of global coverage, we urge the development of in-depth, localized benchmarks tailored to the unique socio-spatial characteristics of diverse regions to foster more responsible and equitable VLMs.

AI Knows Where You Are: Exposure, Bias, and Inference in Multimodal Geolocation with KoreaGEO

Personality is an important concept in psychology that reflects individual differences in thinking and behavior, and has significant applications across various fields. Most existing personality analysis methods address this issue at the bag level, treating the entire corpus gathered from one individual as a single unit for classification. However, this paradigm presents several challenges. From the data perspective, collecting a large corpus for each individual and performing comprehensive annotations pose significant difficulties in both data collection and labeling. On the application side, concentrating on classifying the entire corpus limits its applicability in more common single-instance scenarios. To address these issues, we propose a new task paradigm in text-based personality representation learning. Specifically, we construct a triplet personality trend comparison dataset to learn single-sentence personality embeddings with desirable metric properties. This approach removes the traditional constraints on data sources, facilitating dataset expansion, and can leverage the transfer capabilities of embeddings to easily adapt to various downstream tasks. Our experiments show that the learned embeddings significantly boost performance by a relative 10% across various applications, including personality detection, personality retrieval, and emotion translation prediction. Our dataset and code will be publically available.

Downloads

Next from EMNLP 2025

FB-Bench: A Fine-Grained Multi-Task Benchmark for Evaluating LLMs' Responsiveness to Human Feedback

Similar lecture

Extracting Latent Steering Vectors from Pretrained Language Models