Austria

Inference with modern Large Language Models (LLMs) is both computationally expensive and time-consuming. Speculative decoding has emerged as a promising solution, but existing approaches face key limitations: training-based methods require a draft model that is challenging to obtain and lacks generalizability, while training-free methods offer limited speedup gains. In this work, we present Spectra, a novel framework for accelerating LLM inference without the need for additional training or modification to the original LLM. Spectra introduces two new techniques for efficiently utilizing internal and external speculation, each outperforming corresponding state-of-the-art (SOTA) methods independently. When combined, these techniques achieve up to a 4.08x speedup across various benchmarks and LLM architectures, significantly surpassing existing training-free approaches. The implementation of Spectra is publicly available.

ACL 2025

SPECTRA: Faster Large Language Model Inference with Optimized Internal and External Speculation

training-free

speculative decoding

inference acceleration

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

technical paper

This work aims to understand how scaling improves language models, specifically in terms of training dynamics. We find that language models undergo loss deceleration early in training—an abrupt slowdown in the rate of loss improvement, resulting in piecewise linear behaviour of the loss curve in log-log space. Scaling up the model mitigates this transition by (1) decreasing the loss at which deceleration occurs, and (2) improving the log-log rate of loss improvement after deceleration. We attribute loss deceleration to a type of degenerate training dynamics we term zero-sum learning (ZSL). In ZSL, per-example gradients become systematically opposed, leading to destructive interference in per-example changes in loss. As a result, improving loss on one subset of examples degrades it on another, bottlenecking overall progress. Loss deceleration and ZSL provide new insights into the training dynamics underlying language model scaling laws, and could potentially be targeted directly to improve language models independent of scale. We make our code and artefacts available at: https://github.com/mirandrom/zsl

Understanding Language Model Scaling Laws in Terms of Training Dynamics via Loss Deceleration and Zero-Sum Learning

To facilitate robust and trustworthy deployment of large language models (LLMs), it is essential to quantify the reliability of their generations through uncertainty estimation. While recent efforts have made significant advancements by leveraging the internal logic and linguistic features of LLMs to estimate uncertainty scores, our empirical analysis highlights the pitfalls of these methods to strike a harmonized estimation between indication, balance, and calibration, which hinders their broader capability for accurate uncertainty estimation. To address this challenge, we propose CUE (Corrector for Uncertainty Estimation): A straightforward yet effective method that employs a lightweight model trained on data aligned with the target LLM's performance to adjust uncertainty scores. Comprehensive experiments across diverse models and tasks demonstrate its effectiveness, which achieves consistent improvements of up to 60% over existing methods.

Towards Harmonized Uncertainty Estimation for Large Language Models

Recent advances in large language models (LLMs) have enabled social simulation through multi-agent systems. Prior efforts focus on agent societies created from scratch, assigning agents with newly defined personas. However, simulating established fictional worlds and characters remain largely underexplored, despite its significant practical value. In this paper, we introduce BookWorld, a comprehensive system for constructing and simulating book-based multi-agent societies. BookWorld's design covers comprehensive real-world intricacies, including diverse and dynamic characters, fictional worldviews, geographical constraints and changes, e.t.c. BookWorld enables diverse applications including story generation, interactive games and social simulation, offering novel ways to extend and explore beloved fictional works. Through extensive experiments, we demonstrate that BookWorld generates creative, high-quality stories while maintaining fidelity to the source books, surpassing previous methods with a win rate of 75.36%. The code and demo of this paper can be found at the project page: https://bookworld2025.github.io/.

BOOKWORLD: From Novels to Interactive Agent Societies for Story Creation

Large Language Models (LLMs) need to adapt their predictions to diverse cultural contexts to benefit diverse communities across the world. While previous efforts have focused on single-LLM, single-turn approaches, we propose to exploit the complementary strengths of multiple LLMs to promote cultural adaptability. We introduce a Multi-Agent Debate framework, where two LLM-based agents debate over a cultural scenario and collaboratively reach a final decision. We propose two variants: one where either LLM agents exclusively debate and another where they dynamically choose between self-reflection and debate during their turns. We evaluate these approaches on 7 open-weight LLMs (and 21 LLM combinations) using the NormAd-ETI benchmark for social etiquette norms in 75 countries. Experiments show that debate improves both overall accuracy and cultural group parity over single-LLM baselines. Notably, multi-agent debate enables relatively small LLMs (7-9B) to achieve accuracies comparable to that of a much larger model (27B parameters).

Multiple LLM Agents Debate for Equitable Cultural Alignment

Current Large Language Model (LLM) evaluation frameworks utilize the same static prompt template across all models under evaluation. This differs from the common industry practice of using prompt optimization (PO) techniques to optimize the prompt for each model to maximize application performance. In this paper, we investigate the effect of PO towards LLM evaluations. Our results on public academic and internal industry benchmarks show that PO greatly affects the final ranking of models. This highlights the importance of practitioners performing PO per model when conducting evaluations to choose the best model for a given task.

Optimization before Evaluation: Evaluation with Unoptimized Prompts Can be Misleading

In recent years, large language models have achieved significant success in generative tasks (e.g., speech cloning and audio generation) related to speech, audio, music, and other signal domains. A crucial element of these models is the discrete acoustic codecs, which serve as an intermediate representation replacing the mel-spectrogram. However, there exist several gaps between discrete codecs and downstream speech language models. Specifically, 1) Due to the reconstruction paradigm of the Codec model and the structure of residual vector quantization, the initial channel of the codebooks contains excessive information, making it challenging to directly generate acoustic tokens from weakly supervised signals such as text in downstream tasks. 2) Achieving good reconstruction performance requires the utilization of numerous codebooks, which increases the burden on downstream speech language models. Consequently, leveraging the characteristics of speech language models, we propose Language-Codec. In the Language-Codec, we introduce a Masked Channel Residual Vector Quantization (MCRVQ) mechanism along with improved fourier transform structures, refined discriminator design to address the aforementioned gaps. We compare our method with competing audio compression algorithms and observe significant outperformance across extensive evaluations. Furthermore, we also validate the efficiency of the Language-Codec on downstream speech language models. The source code and pretrained models will be open-sourced after the paper is accepted. Codes are available at https://github.com/jishengpeng/Languagecodec.

Bridging Discrete Codec Representations and Speech Language Models

Implicit hate speech has become a significant challenge for online platforms, as it often avoids detection by large language models (LLMs) due to its indirectly expressed hateful intent. This study identifies the limitations of LLMs in detecting implicit hate speech, particularly when disguised as seemingly harmless expressions in a rhetorical device. To address this challenge, we employ a Jailbreaking strategy and Energy-based Constrained Decoding techniques, and design a small model for measuring the energy of metaphorical rhetoric. This approach can lead to LLMs generating metaphorical implicit hate speech. Our research reveals that advanced LLMs, like GPT-4o, frequently misinterpret metaphorical implicit hate speech, and fail to prevent its propagation effectively. Even specialized models, like ShieldGemma and LlamaGuard, demonstrate inadequacies in blocking such content, often misclassifying it as harmless speech. This work points out the vulnerability of current LLMs to implicit hate speech, and emphasizes the improvements to address hate speech threats better.

Sheep's Skin, Wolf's Deeds: Are LLMs Ready for Metaphorical Implicit Hate Speech?

Entity disambiguation (ED) is the task of linking mentions in text to corresponding entries in a knowledge base. Dual Encoders address this by embedding mentions and label candidates in a shared embedding space and applying a similarity metric to predict the correct label. In this work, we focus on evaluating key design decisions for Dual Encoder-based ED, such as its loss function, similarity metric, label verbalization format, and negative sampling strategy. We present the resulting model VerbalizED, a document-level Dual Encoder model that includes contextual label verbalizations and efficient hard negative sampling. Additionally, we explore an iterative prediction variant that aims to improve the disambiguation of challenging data points. To support our analysis, we first conduct comprehensive ablation experiments on specific design decisions using AIDA-Yago, followed by large-scale, multi-domain evaluation on the ZELDA benchmark.

Evaluating Design Decisions for Dual Encoder-based Entity Disambiguation

We present a family of encodings for sequence labeling dependency parsing, based on the concept of hierarchical bracketing. We show that the existing 4-bit projective encoding belongs to this family, but it is suboptimal in the number of labels used to encode a tree. We derive an optimal hierarchical bracketing, which minimizes the number of symbols used and encodes projective trees using only 12 distinct labels (vs. 16 for the 4-bit encoding). We also extend optimal hierarchical bracketing to support arbitrary non-projectivity in a more compact way than previous encodings. Our new encodings yield competitive accuracy on a diverse set of treebanks.

Hierarchical Bracketing Encodings for Dependency Parsing as Tagging

This paper introduces BMIKE-53, a comprehensive benchmark for cross-lingual in-context knowledge editing (IKE), spanning 53 languages and three KE datasets: zsRE, CounterFact, and WikiFactDiff. Cross-lingual KE, which requires knowledge edited in one language to generalize across diverse languages while preserving unrelated knowledge, remains underexplored. To address this, we systematically evaluate IKE under zero-shot, one-shot, and few-shot setups, including tailored metric-specific demonstrations. Our findings reveal that model scale and demonstration alignment critically govern cross-lingual editing efficacy, with larger models and tailored demonstrations significantly improving performance. Linguistic properties, particularly script type, strongly influence outcomes, with non-Latin languages underperforming due to issues like language confusion.

Downloads

Next from ACL 2025

Understanding Language Model Scaling Laws in Terms of Training Dynamics via Loss Deceleration and Zero-Sum Learning

Similar lecture

A Drop-In Solution for On-the-Fly Adaptation of Speculative Decoding in Large Language Models