United States

The pretraining data mixture of Large Language Models (LLMs) constitutes their &quot;digital DNA&quot;, shaping model behaviors, capabilities, and failure modes. Yet this composition is rarely disclosed, making post-hoc auditing of data combination or provenance difficult. In this work, we formalize $\textbf{\texttt{Data Mixture Surgery (DMS)}}$: given only generated text from a target LLM, estimate the domain-level distribution of its pretraining corpus under a predefined taxonomy. We propose $\textbf{\texttt{LLMSurgeon}}$, a strong framework that casts DMS as an inverse problem under the label-shift assumption. Rather than directly aggregating classifier outputs, $\texttt{LLMSurgeon}$ estimates a calibrated $\textit{soft}$ confusion matrix and solves a constrained inverse problem to correct systematic domain confusion and recover the latent mixture prior. To evaluate, we introduce $\textbf{\texttt{LLMScan}}$, a recipe-verifiable evaluation suite built from open-source LLMs with transparent pretraining mixtures. Across LLMScan, $\texttt{LLMSurgeon}$ recovers domain mixtures with high fidelity under fixed protocols. Our work presents a practical, post-hoc approach for auditing the $\textit{digital DNA}$ of foundation models without access to their training data. Code is available at: https://github.com/Yaxin9Luo/LLMSurgeon.

ACL 2026

LLMSurgeon: Diagnosing Data Mixture of Large Language Models

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

The pretraining data mixture of Large Language Models (LLMs) constitutes their "digital DNA", shaping model behaviors, capabilities, and failure modes. Yet this composition is rarely disclosed, making post-hoc auditing of data combination or provenance difficult. In this work, we formalize $\textbf{\texttt{Data Mixture Surgery (DMS)}}$: given only generated text from a target LLM, estimate the domain-level distribution of its pretraining corpus under a predefined taxonomy. We propose $\textbf{\texttt{LLMSurgeon}}$, a strong framework that casts DMS as an inverse problem under the label-shift assumption. Rather than directly aggregating classifier outputs, $\texttt{LLMSurgeon}$ estimates a calibrated $\textit{soft}$ confusion matrix and solves a constrained inverse problem to correct systematic domain confusion and recover the latent mixture prior. To evaluate, we introduce $\textbf{\texttt{LLMScan}}$, a recipe-verifiable evaluation suite built from open-source LLMs with transparent pretraining mixtures. Across LLMScan, $\texttt{LLMSurgeon}$ recovers domain mixtures with high fidelity under fixed protocols. Our work presents a practical, post-hoc approach for auditing the $\textit{digital DNA}$ of foundation models without access to their training data. Code is available at: https://github.com/Yaxin9Luo/LLMSurgeon.

technical paper

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. 
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) 
<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>

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 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.

Large language models (LLMs) have progressed rapidly in complex reasoning and question answering, yet LLM hallucination remains a central bottleneck that hinders practical deployment, especially for commercial black-box LLMs accessible only via APIs. Existing uncertainty quantification methods typically depend on computationally expensive multiple sampling or internal parameters, which prevents real-time estimation and fails to capture information implicit in the black-box reasoning process. To address this issue, we propose Distribution-Aligned Adversarial Distillation (DisAAD), which introduces a generation-discrimination architecture to guide a lightweight proxy model to learn the high-quality regions of the output distribution of the black-box LLM, thus effectively endowing it with the ability to "know whether the black-box LLM knows or not". Subsequently, we use the proxy model to reproduce the specific responses of the black-box LLM and estimate the corresponding uncertainty based on evidence learning. Extensive experiments have verified the effectiveness and promise of our proposed method, indicating that a proxy model even one that only accounts for $1$\% of the target LLM’s size can achieve reliable uncertainty quantification.

Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation

Current large language models (LLMs) often exhibit performance imbalances between dominant languages (e.g., English) and non-dominant ones due to the skewed distribution of pretraining data. A common strategy to address this issue is to enhance cross-lingual alignment, thereby facilitating non-dominant language processing. However, existing methods typically rely on additional training objectives or language-specific parameters, which increase training complexity and cost. In this work, we propose a selective bidirectional language projection framework that enables efficient multilingual alignment and language shift using the intrinsic parameters. Specifically, we first identify the layers most sensitive to language projection between non-dominant and dominant languages through neuron activation analysis. We then perform sequential language projection within the selected layers by mapping non-dominant representations into the dominant language space and reverting them before generation. The bidirectional projection benefits the subsequent instruction tuning in non-dominant languages. Experiments on seven benchmarks demonstrate that our method remarkably enhances the performance of non-dominant languages. Further analyses indicate that our method learns better internal representations and exhibits strong generalization capabilities.

SiLP: Enhancing Non-Dominant Language Capabilities with a Selective Bidirectional Language Projection Framework

Reinforcement learning has proven effective for enhancing multi-step reasoning in Large Language Models (LLMs), yet its benefits have not fully translated to multilingual contexts. Existing methods struggle with a fundamental trade-off: prioritizing input-language consistency severely hampers reasoning quality, while prioritizing reasoning often leads to unintended language drift toward English. We address this challenge with LANG, a novel framework that leverages language-conditioned hints to guide exploration in non-English reasoning tasks. Our method incorporates two key mechanisms to prevent dependency on these hints: a progressive decay schedule that gradually withdraws scaffolding, and a language-adaptive switch that tailors learning horizons to specific language difficulties. Empirical results on challenging multilingual mathematical benchmarks reveal that LANG substantially enhances reasoning performance without compromising language consistency. Moreover, we show that our framework generalizes beyond mathematics, fostering more consistent language alignment across model layers.

LANG: Reinforcement Learning for Multilingual Reasoning with Language-Adaptive Hint Guidance

Evolution, the engine behind the survival and growth of life on Earth, operates through the population-based process of reproduction. Inspired by this principle, this paper formally defines a newly emerging problem: the population-based evolution of large language models (LLMs). We introduce a novel framework that starts with a population of parent LLMs and allows this population to evolve through four key operations: (i) crossover, merging the weights of different parents to create offspring LLMs, (ii) mutation, introducing small, random changes to model weights to foster diversity, (iii) selection, prioritizing high-performing models, and (iv) succession, transferring the learned experience from parent to offspring LLMs. With only 200 samples per new task, the LLM population evolves rapidly to adapt to the task at hand, without any gradients. Experiments on 12 datasets show that our framework consistently outperforms existing multi-LLM merging and adaptation methods, achieving relative performance gains of up to 54.8 over the best LLM in the initial population. Moreover, our framework allows for (i) the evolution of LLMs across multiple new tasks simultaneously, (ii) scaling effectively with populations of up to 40 LLMs, and (iii) even zero-shot generalization to unseen held-out tasks. Code: https://github.com/ZhangYiqun018/GENOME

Nature-Inspired Population-Based Evolution of Large Language Models

Multi-turn, long-horizon tasks are increasingly common for large language models (LLMs), but solving them typically requires many sequential model invocations, accumulating substantial inference costs. Here, we study cost-aware multi-turn LLM routing: selecting which model to invoke at each turn from a model pool, given a fixed cost budget. We propose MTRouter, which encodes the interaction history and candidate models into joint history–model embeddings, and learns an outcome estimator from logged trajectories to predict turn-level model utility. Experiments show that MTRouter improves the performance–cost trade-off: on ScienceWorld, it surpasses GPT-5 while reducing total cost by 58.7\%; on Humanity’s Last Exam (HLE), it achieves competitive accuracy while reducing total cost by 43.4\% relative to GPT-5, and these gains even carry over to held-out tasks. Further analyses reveal several mechanisms underlying its effectiveness: relative to prior multi-turn routers, MTRouter makes fewer model switches, is more tolerant to transient errors, and exhibits emergent specialization across models.
Code: https://github.com/ZhangYiqun018/MTRouter

MTRouter: Cost-Aware Multi-Turn LLM Routing with History–Model Joint Embeddings

Logical reasoning is a key capability of large language models, yet current benchmarks focus almost entirely on tasks that just check basic logical consistency and overlook the reflective reasoning required for paradox detection and resolution. To fill the gap, we present ParaSuite, the first pipeline dedicated to paradox research that automates data synthesis, evaluation, and training. We introduce PARADOX, a synthetic, high-quality data spanning two difficulty tiers and three academic domains, accompanied by specialized evaluation metrics and solving algorithms. We propose ParadoxBreaker-7B, trained with Mutual-Information Guided Fine-Tuning and reinforcement learning step verify paradox reward(PAPO). Experiments demonstrate significant improvements in both paradoxical and general STEM reasoning.

ParaSuite: Boosting LLM Reasoning via Paradox Resolution

Existing datasets for evaluating text-to-image generation focus mostly on real-life images, which poses challenges for assessing academic
figure generation given real scientific captions, which is a hot topic in AI for Science. To fill the gap, we propose HE4AFG, a novel dataset
which first provides a Holistic Evaluation for Academic caption-to-Figure Generation (AFG). Specifically, HE4AFG collects real figure captions from 8 scientific domains and finally generates 3,900 evaluation samples (particularly, including multi-panel figures) using 5 mainstream large multimodal models (LMMs). For each sample, we provide high-quality human ratings in terms of three aspects—scientific aesthetic (SA), topic relevance (TR), and attribute correctness (AC). Moreover, we present two trainable models: (1) HE4AFG-E, an automated Evaluation model for AFG, which generates aspect-aware training examples and then use them to train three aspect-specific evaluation modules via contrastive learning; (2) HE4AFG-R, an automated Refinement model, which generates and utilizes feedback on the quality of the figures (e.g., unfaithful elements) to continuously improve AFG. Extensive experiments on HE4AFG demonstrate the effectiveness and performance advantages of our models.

Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs

Reinforcement Learning with Verifiable Rewards (RLVR) improves the reasoning capability of Large Language Models (LLMs). Current RLVR trains LLMs on all generated tokens, rather than exploring which tokens actually contribute to reasoning. We propose AIPO(Adaptive--Information Policy Optimization), which focuses updates on those decisive tokens discovered on the fly. AIPO estimates each hidden state’s mutual information to score tokens. Policy gradients are then computed only on these critical tokens, using an advantage that blends information gain and verifiable correctness. To improve the efficiency of mutual-information estimation, AIPO adopts a Random–Fourier approximation of the Hilbert–Schmidt Independence Criterion. Across five math and science benchmarks, AIPO yields up to +20% accuracy over strong RLVR baselines while updating merely 10% of tokens, demonstrating superior efficiency and effectiveness. Our findings highlight the importance of information–driven token selection for efficient and effective reinforcement learning of LLM reasoning.

AIPO: Adaptive Information Guided Token-Level Reinforcement Learning for Large Language Model Reasoning

Cross-lingual learning enables the transfer of structured sentiment knowledge from high-resource languages to unlabeled or low-resource languages, but prior work has largely focused on coarse-grained sentiment classification or aspect extraction. In contrast, zero-shot cross-lingual aspect--opinion--sentiment triplet extraction (ASTE), which extracts sentiment triplets of the form $(\textit{aspect term}, \textit{opinion term}, \textit{sentiment polarity})$, remains underexplored. We propose a unified framework that leverages large language models (LLMs) as both structured pseudo-label generators and semantic teachers for ASTE. Our approach employs stepwise structured prompting over aspect- and opinion-aware code-switched variants to generate reliable pseudo triplets, followed by a multi-variant consistency filter to retain high-confidence supervision. We further introduce a triplet-aware contrastive distillation objective that aligns student triplet representations with LLM-encoded semantic embeddings. During inference, only the student ASTE model is used, without requiring LLM access. Experiments on four non-Indic and four low-resource Indic target languages show consistent improvements over strong cross-lingual and LLM-based baselines. The proposed method yields an absolute micro-F1 improvement of 5.3 points on non-Indic languages and 3.8 points on low-resource Indic languages compared to the best competing approach. Ablation results further validate the complementary roles of aspect- and opinion-aware code-switched prompting and triplet-aware contrastive distillation, with larger relative gains observed in low-resource Indic settings.

CLAOCS-TX: Cross-Lingual Triplet Extraction with Aspect-Opinion-Aware Code-Switched Prompting and LLM-Guided Contrastive Distillation

Mobile GUI agents powered by LMMs can perceive screens and follow instructions, yet existing benchmarks largely target short, linear workflows and step-level accuracy, offering limited insight into long-horizon planning and decision-making under branching structures. We present DAC-Bench, a decision-aware benchmark with compositional tasks comprising 830 episodes and 11,345 action steps across 35 applications on Android and iOS. Tasks are organized into Sequential, Conjunctive, Conditional, and Hierarchical structures, reflecting real-world multi-step and branching interaction patterns. To complement standard step-level evaluation, we introduce weighted longest common subsequence to capture length-sensitive progress and decision accuracy for branch correctness. Evaluations across 7 diverse agents show substantial performance degradation compared to prior benchmarks, with success rates dropping below 5% on 6–8 step tasks and branch accuracy averaging 38%, highlighting challenges in conditional decision-making. By exposing these failure modes, DAC-Bench provides a challenging and diagnostic benchmark for advancing decision-aware mobile GUI agents. Our code and dataset are available at: https://github.com/YuqingZhangMirror12/DAC-Bench.

Downloads

Next from ACL 2026

Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation

Similar lecture

Generalists vs. Specialists: Evaluating Large Language Models for Urdu