United States

Diffusion language models generate text by iteratively denoising all tokens in parallel, but when and where their hidden states encode whether the output will be functionally correct remains unknown. We present the first probing study of DLM internals, training linear classifiers on hidden states extracted at intermediate denoising steps to predict functional correctness. Across two models (LLaDA-8B, Dream-7B) and four tasks (JSON schema validation, math reasoning, code generation, science QA), we find that (1) correctness signal is already present at step 0, before any denoising occurs, (2) signal emergence is task-dependent, with structural tasks showing flat profiles and reasoning tasks showing gradual buildup, and (3) layer dynamics diverge across architectures. We further demonstrate that per-step probe confidence can identify instances likely to fail, enabling selective generation that avoids 36-98% of wasted compute depending on the task.

ACL 2026

Probing Functional Correctness in Diffusion Language Models

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields 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.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>
<br><br><br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

Registration Required

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

technical paper

Tokenizer design is a foundational but underexamined component of large language model (LLM) development. We introduce TokLens, an open-source toolkit for evaluating tokenizer quality across languages using six intrinsic metrics: fertility, characters per token, compression ratio, normalized sequence length, single-token retention rate, and cross-lingual parity. We evaluate 24 tokenizers from major LLM families across 15 typologically diverse languages and correlate intrinsic metrics with downstream performance on the Open LLM Leaderboard v2. Our analysis reveals stark cross-lingual disparities: GPT-2's tokenizer produces 56x more tokens per word in Japanese than in English, while newer tokenizers like Qwen2.5 and Gemma-2 reduce this gap to under 4x. No intrinsic metric predicts English benchmark performance after controlling for model size. However, when we extend the analysis to multilingual benchmarks (MMLU-ProX) using linear mixed-effects models that account for repeated measures within models, tokenizer metrics significantly predict per-language performance (STRR: $\beta = +5.7$, $z = 18.5$, $p < 0.001$). A controlled experiment with the Qwen2.5 family, which shares a single tokenizer across five model sizes, reveals that languages with higher single-token retention rate benefit more from scaling ($\rho = 0.91$, $p < 0.001$). These results suggest that tokenizer quality has a measurable impact on multilingual LLM performance, even when model architecture and training pipeline are held constant.

TokLens: A Multilingual Lens on Tokenizer Quality for LLMs

Large language models are increasingly used as evaluators, yet recent work shows that their judgments can be distorted by superficial properties of responses, including epistemic markers such as maybe, probably, and I think. Prior evidence has suggested that hedged but correct answers may be penalized by LLM judges, but it remains unclear whether this effect survives in multilingual and code-mixed settings once benchmark construction artifacts are controlled. We present PolyJudge-Uncertain, a controlled benchmark for multilingual judge robustness under hedging, built from 320 base question-answer items and expanded into 5,120 examples across English, Hindi, Hinglish, and Bengali, with balanced assertive versus hedged formulations and correct versus incorrect answers. We evaluate a reference-guided judge using both pointwise and pairwise protocols. In an initial pilot, we observe a large pointwise penalty against hedged answers. However, after naturalizing templates and introducing template-quality control, that pointwise effect largely disappears: pointwise accuracy reaches 99.8\% overall, with no meaningful assertive--hedged gap. The robust remaining effect emerges in pairwise comparison. When two answers are equally correct and differ only in style, the judge prefers the assertive response in 1,276 of 1,280 cases. These results refine the current narrative on uncertainty bias in LLM-as-a-judge. In our setting, the strongest stable failure mode is not pointwise misclassification of hedged truth, but pairwise over-preference for assertiveness among semantically equivalent outputs. More broadly, our findings show that multilingual judge-bias claims are highly sensitive to template naturalness and benchmark construction, making benchmark auditing a first-class methodological requirement for this line of work.

Confidence as a Tie-Breaker: Reassessing Multilingual Hedging Bias in LLM-as-a-Judge Evaluation

As large language models (LLMs) grow, their compute and memory demands become prohibitive for on-device deployment. Quantization is a crucial technique to shrink model footprint and accelerate inference, but aggressively low-bit weight-activation quantization schemes often sacrifice accuracy. Quantization Aware Training (QAT) is a commonly used paradigm to minimize quantization noise, but is extremely expensive to train and often unscalable to large models. We introduce PE-QAT, a parameter-efficient framework targeting per-channel 4-bit weight-activation quantization of LLMs, which aims to preserve model accuracy while significantly reducing resource requirements. The proposed method freezes the base model and trains lightweight LoRA adapters by fake quantizing the merged-weight model, enabling PE-QAT to scale efficiently unlike full QAT. We apply fake quantization with Straight-Through Estimators (STE) to the merged weights, allowing the adapters to explicitly compensate for quantization noise during training. One of the biggest challenges with quantizing activations alongside weights is addressing outliers that are orders of magnitude larger than other activations, which inflate quantization scales and suppress lower-magnitude values. To mitigate the impact of severe activation outliers, PE-QAT jointly learns per-channel smoothing factors and symmetric activation clipping thresholds. PE-QAT retains accuracy within 0.11 percentage points of the full-precision baseline on Llama-2-7B zero-shot tasks while training only 1.26% of total parameters.

PE-QAT: Parameter-Efficient Quantization Aware Training for Large Language Models

Evaluating the grammatical abilities of large language models (LLMs) is important for both NLP and linguistic theory. We investigate the ability of large language models (LLMs) to perform acceptability judgments in a forced-choice paradigm. We evaluate a subset of LLMs on 150 minimal sentence pairs sampled from Linguistic Inquiry and categorized using BLiMP linguistic phenomena. Our results show that while LLMs approximate human judgments, performance varies across models and phenomenon types, with less stable behavior on more complex phenomena. Prompting strategies have minimal impact.

Linguistically-Informed Evaluation of LLMs on Acceptability Judgments in a Forced-Choice Paradigm

Text-to-SQL technology has evolved rapidly, with diverse academic methods achieving impressive results. However, deploying these techniques in real-world systems remains challenging due to limited integration tools. Despite these advances, we introduce Squrve, a unified, modular, and extensive Text-to-SQL framework designed to bring together research advances and real-world applications. Squrve first establishes a universal execution paradigm that standardizes invocation interfaces, then proposes a multi-actor collaboration mechanism based on seven abstracted effective atomic actor components. Experiments on widely adopted benchmarks demonstrate that the collaborative workflows consistently outperform the original individual methods, thereby opening up a new effective avenue for tackling complex real-world queries. The codes are available at https://anonymous.4open.science/r/Squrve-Annoy-E5B4.

Squrve: A Unified and Modular Framework for Complex Real-World Text-to-SQL Tasks

Building retrieval-augmented generation (RAG) systems often requires
combining separate tools for retrieval, re-ranking, and generation,
with incompatible data formats, evaluation pipelines, and deployment workflows.
We present \framework, an open-source Python toolkit that unifies these stages
in a single modular framework.\footnote{PyPI: \url{https://pypi.org/project/rankify/}}
\textsuperscript{,}\footnote{GitHub: \url{https://github.com/DataScienceUIBK/Rankify}}
\textsuperscript{,}\footnote{Docs: \url{https://rankify.readthedocs.io}}%
\textsuperscript{,}\footnote{Video: \url{https://youtu.be/kkLzomrM2ec}}
\framework provides 42 benchmark datasets with pre-retrieved documents and
pre-built indices, 15 retrievers (sparse, dense, and reasoning-augmented),
and 24 re-ranking models spanning 41 pointwise, pairwise, and listwise variants.
It also supports \textbf{6} RAG strategies across four inference backends
(Hugging Face, vLLM, LiteLLM, and OpenAI), enabling consistent experimentation
from local models to hosted APIs.
A unified pipeline interface allows users to compose retrieve--rerank--generate
workflows in a few lines of code, while an agentic assistant (\texttt{RankifyAgent}), a
REST server (\texttt{RankifyServer}), and an interactive web
playground support deployment and non-programmatic exploration.
Across 200+ configurations on QA and BEIR/TREC benchmarks with six generator LLMs,
re-ranking consistently improves downstream performance, yielding gains of
5--15 points in Exact Match and up to 8.5 points in RAGAS context precision
across diverse retriever--generator combinations.

Rankify: A Comprehensive Python Toolkit for Retrieval, Re-Ranking, and Retrieval-Augmented Generation

Synthetic data has emerged as a crucial solution to the data scarcity bottleneck in large language models (LLMs), particularly for specialized domains and low-resource languages. However, the broader adoption of existing synthetic data tools is severely hindered by convoluted workflows, fragmented data standards, and limited scalability across modalities.
To address these limitations, we develop DataArc-SynData-Toolkit, an open-source framework featuring: (1) a configuration-driven, end-to-end pipeline equipped with an intuitive visual interface and simplified CLI for exceptional usability; (2) a unified, quality-controllable synthesis paradigm that standardizes multi-source data generation to ensure high reusability; and (3) a highly modular architecture designed for seamless multimodal, multilingual, and multi-task adaptation.
We apply the toolkit in multiple application scenarios. 
Experimental results demonstrate that our toolkit achieves an optimal balance between generation efficiency and data quality. By offering an end-to-end and visually interactive pipeline, DataArc-SynData-Toolkit significantly lowers the technical barrier to synthetic data generation and subsequent model training, accelerating its practical deployment in real-world applications.

DataArc-SynData-Toolkit: A Unified Closed-Loop Framework for Multi-Path, Multimodal, and Multilingual Data Synthesis

Recent work in natural language processing, human-computer interaction, and computational social science takes interest in the study of decentralized content moderation, in which individual communities largely determine their own norms, rules, and enforcement thereof. A key challenge to this body of work is that, once moderated, content and related variables become difficult or impossible to recover; previous work often relied on 3rd-party historical data sources, but recent world events, legal disputes, and policy shifts have significantly disrupted these services, practically disabling their research use-cases. Therefore, in order to conduct new research and reproduce previous results, researchers must record content as it's created, and monitor variables of interest over time. In this work we present and publicly release a software system for the dynamic monitoring of Reddit posts, communities, and moderation actions, to enable scaleable and reproducible research on decentralized platform governance and content moderation. To the authors' knowledge, this system is presently the only available solution for general-purpose, real-time, policy-compliant longitudinal data collection on Reddit. Furthermore, the system's integration with the official Reddit API enables the collection of authentication-gated data such as community engagement metrics and moderation team information, which was impossible in previous historical data sources.

A System for Dynamically Tracking Content Moderation on Reddit

Legal practitioners in Thailand must navigate fragmented government websites to research over 3,800 active laws and 87,000 Supreme Court decisions, with no unified tool for cross-referencing, version tracking, or structural navigation. We present FourCorners, a deployed platform that addresses five practitioner pain points through three modules built on a temporal legal knowledge graph covering 552K nodes and 6.3M edges: (1) an AI legal assistant that performs grounded generation via structured graph retrieval, streaming verified source content inline with responses; (2) an interactive law reader that translates graph structure into navigation and comparison features; and (3) a court decision explorer with version-aware citations produced by temporal entity resolution across 87,394 rulings. The system discovers implicit cross-corpus relationships through co-citation analysis of Supreme Court decisions. Interviews with 20 legal professionals reveal that inline source verification fundamentally changes how practitioners interact with AI-generated legal content, and that cross-corpus enrichment surfaces legal relationships that existing tools leave invisible.

FourCorners: Grounded Thai Legal Research over a Temporal Knowledge Graph

The development of audio foundation models has accelerated rapidly since the emergence of GPT-4o. However, the lack of comprehensive evaluation has become a critical bottleneck for further progress in the field, particularly in audio generation. Current audio evaluation faces three major challenges: 
(1) audio evaluation lacks a unified framework, with datasets and code scattered across various sources;
(2) audio codec, as a key component of audio foundation models, lacks a widely accepted and holistic evaluation methodology; 
(3) existing speech benchmarks are heavily reliant on English, making it challenging to objectively assess models' performance on Chinese.
We introduce UltraEval-Audio, a unified framework addressing these challenges through a modular architecture supporting 10 languages, 14 task categories, 24 models, and 36 benchmarks with one-command evaluation and real-time leaderboards. For audio codec, we propose a three-dimensional evaluation scheme covering semantic accuracy, timbre fidelity, and acoustic quality. For Chinese evaluation, we introduce two new benchmarks: SpeechCMMLU and SpeechHSK. Our code, benchmarks, and leaderboards are available at https://github.com/OpenBMB/UltraEval-Audio.

Downloads

Next from ACL 2026

TokLens: A Multilingual Lens on Tokenizer Quality for LLMs

Similar lecture

Advancing Reasoning in Diffusion Language Models with Denoising Process Rewards

Stay up to date with the latest Underline news!

PRODUCT

COMPANY

RESOURCES

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from ACL 2026

TokLens: A Multilingual Lens on Tokenizer Quality for LLMs

Similar lecture

Advancing Reasoning in Diffusion Language Models with Denoising Process Rewards

Stay up to date with the latest Underline news!

PRODUCT

COMPANY

RESOURCES

Downloads