Austria

Current self-correction approaches in text-to-SQL face two critical limitations: 1) Conventional self-correction methods rely on recursive self-calls of LLMs, resulting in multiplicative computational overhead, and 2) LLMs struggle to implement effective error detection and correction for monolithic SQL queries, as they fail to demonstrate the underlying reasoning path. In this work, we propose **SHARE**, a **S**LM-based **H**ierarchical **A**ction cor**RE**ction assistant that enables LLMs to perform more precise error localization and efficient correction. SHARE orchestrates three specialized Small Language Models (SLMs) in a sequential pipeline, where it first transforms monolithic SQL queries into stepwise action trajectories that reveal underlying reasoning, followed by a two-phase granular refinement. We further propose a novel hierarchical self-evolution strategy for data-efficient training. Our experimental results demonstrate that SHARE effectively enhances self-correction capabilities while proving robust across various LLMs. Furthermore, our comprehensive analysis shows that SHARE maintains strong performance even in low-resource training settings, which is particularly valuable for text-to-SQL applications with data privacy constraints.

ACL 2025

SHARE: An SLM-based Hierarchical Action CorREction Assistant for Text-to-SQL

self-correction

text-to-sql

semantic parsing

## 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*

EMNLP 2025

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

poster

### 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)

The evolution of Large Language Models (LLMs) has underscored the necessity for benchmarks designed for various languages and cultural contexts. To address this need for Vietnamese, we present the first Vietnamese Multitask Language Understanding (VMLU) Benchmarks. The VMLU benchmarks consist of four datasets that assess different capabilities of LLMs, including general knowledge, reading comprehension, reasoning, and conversational skills. This paper also provides an insightful overview of the current state of some dominant LLMs, such as Llama-3, Qwen2.5, and GPT-4, highlighting their performances and limitations when measured against these benchmarks. Furthermore, we provide insights into how prompt design can influence VMLU's evaluation outcomes, as well as suggest that open-source LLMs can serve as effective, cost-efficient evaluators within the Vietnamese context. By offering a comprehensive and accessible benchmarking framework, the VMLU Benchmarks aim to foster the development and fine-tuning of Vietnamese LLMs, thereby establishing a foundation for their practical applications in language-specific domains.

VMLU Benchmarks: A comprehensive benchmark toolkit for Vietnamese LLMs

The Transformer architecture has become the standard LLM architecture due to its powerful self-attention mechanism. However, it suffers from quadratic computational complexity and linear memory complexity. RNN-based LLMs have been proposed as alternatives. Yet, RNN models struggle in long-context scenarios, making it challenging to replace self-attention with RNNs. We identify the state size as a critical bottleneck, which is significantly smaller than that of Transformers with a basic context length of 2k. However, simply increasing the state size significantly raises the number of parameters and lowers training efficiency. In this paper, we propose an efficient scaling method to scale the state size of RNN models to match the 2k context length of Transformers, with small parameters overhead. Experimental results demonstrate that scaling the state size significantly enhances long-context understanding. Retrieval performance scales almost linearly with state size, with a 454M model featuring an expanded state achieving performance comparable to a 1.47B model on FDA, a recall-intensive task. These findings highlight state scaling as a promising approach for advancing RNN-based LLMs.

Scaling Laws for RNN LLM in Long-Context Scenarios with State Size

Most progress in recent coder models has been driven by supervised fine-tuning (SFT), while the potential of reinforcement learning (RL) remains largely unexplored, primarily due to the lack of reliable reward data/model in the code domain. In this paper, we address this challenge by leveraging automated large-scale test-case synthesis to enhance code model training. Specifically, we design a pipeline that generates extensive (question, test-cases) pairs from existing code data. Using these test cases, we construct preference pairs based on pass rates over sampled programs to train reward models with Bradley-Terry loss. It shows an average of 10-point improvement for Llama-3.1-8B-Ins and 5-point improvement for Qwen2.5-Coder-7B-Ins through best-of-32 sampling, making the 7B model on par with 236B DeepSeek-V2.5. Furthermore, we conduct reinforcement learning with both reward models and test-case pass rewards, leading to consistent improvements across HumanEval, MBPP, BigCodeBench, and LiveCodeBench (V4). Notably, we follow the R1-style training to start from Qwen2.5-Coder-base directly and show that our RL training can improve model on HumanEval-plus by over 25% and MBPP-plus by 6% for merely 80 optimization steps. We believe our results highlight the huge potential of reinforcement learning in coder models.

ACECODER: Acing Coder RL via Automated Test-Case Synthesis

Chain-of-Thought (CoT) prompting has become the de facto method to elicit reasoning capabilities from large language models (LLMs). However, to mitigate hallucinations in CoT that are notoriously difficult to detect, current methods such as process reward models (PRMs) or self-consistency operate as opaque boxes and do not provide checkable evidence for their judgments, possibly limiting their effectiveness. To address this issue, we draw inspiration from the idea that "the gold standard for supporting a mathematical claim is to provide a proof". We propose a retrospective, step-aware formal verification framework Safe. Rather than assigning arbitrary scores, we strive to articulate mathematical claims in formal mathematical language Lean 4 at each reasoning step and provide formal proofs to identify hallucinations. We evaluate our framework Safe across multiple language models and various mathematical datasets, demonstrating a significant performance improvement while offering interpretable and verifiable evidence. We also propose FormalStep as a benchmark for step correctness theorem proving with $30,809$ formal statements. To the best of our knowledge, our work represents the first endeavor to utilize formal mathematical language Lean 4 for verifying content generated by LLMs, aligning with the reason why formal mathematical languages were created in the first place: to provide a robust foundation for hallucination-prone human-written proofs.

Safe: Enhancing Mathematical Reasoning in Large Language Models via Retrospective Step-aware Formal Verification

Large language models (LLMs) have the potential to revolutionize smart home assistants by enhancing their ability to accurately understand user needs and respond appropriately, which is extremely beneficial for building a smarter home environment. While recent studies have explored integrating LLMs into smart home systems, they primarily focus on handling straightforward, valid single-device operation instructions. However, real-world scenarios are far more complex and often involve users issuing invalid instructions or controlling multiple devices simultaneously. These have two main challenges: LLMs must accurately identify and rectify errors in user instructions and execute multiple user instructions perfectly. To address these challenges and advance the development of LLM-based smart home assistants, we introduce HomeBench, the first smart home dataset with valid and invalid instructions across single and multiple devices in this paper. We have experimental results on 13 distinct LLMs; e.g., GPT-4o achieves only a 0.0\% success rate in the scenario of invalid multi-device instructions, revealing that the existing state-of-the-art LLMs still cannot perform well in this situation even with the help of in-context learning, retrieval-augmented generation, and fine-tuning. Our code and dataset are publicly available at https://github.com/BITHLP/HomeBench.

HomeBench: Evaluating LLMs in Smart Homes with Valid and Invalid Instructions Across Single and Multiple Devices

Iterative data generation and model retraining are widely used to align large language models (LLMs).
It typically involves a policy model to generate on-policy responses and a reward model to guide training data selection. 
Direct Preference Optimization (DPO) further enhances this process by constructing preference pairs of chosen and rejected responses. 
In this work, we aim to \emph{scale up} the number of on-policy samples via repeated random sampling to improve alignment performance. 
Conventional practice selects the sample with the highest reward as chosen and the lowest as rejected for DPO. 
However, our experiments reveal that this strategy leads to a \emph{decline} in performance as the sample size increases. 
To address this, we investigate preference data construction through the lens of underlying normal distribution of sample rewards. 
We categorize the reward space into seven representative points and systematically explore all 21 ($C_7^2$) pairwise combinations. 
Through evaluations on four models using AlpacaEval 2, we find that selecting the rejected response at reward position $\mu - 2\sigma$ rather than the minimum reward, is crucial for optimal performance. 
We finally introduce a scalable preference data construction strategy that consistently enhances model performance as the sample scale increases.

Finding the Sweet Spot: Preference Data Construction for Scaling Preference Optimization

The strong performance of large language models (LLMs) raises extensive discussion on their application to code generation. Recent research suggests continuous program refinements through visible tests to improve code generation accuracy in LLMs. However, these methods suffer from LLMs' inefficiency and limited reasoning capacity. In this work, we propose an LLM programming workflow (LPW) designed to improve both initial code generation and subsequent refinements within a structured two-phase workflow. Specifically, the solution generation phase formulates a solution plan, which is then verified through visible tests to specify the intended natural language solution. Subsequently, the code implementation phase drafts an initial code according to the solution plan and its verification. If the generated code fails the visible tests, the plan verification serves as the intended solution to consistently inform the refinement process for correcting bugs. Compared to state-of-the-art methods across various existing LLMs, LPW significantly improves the Pass@1 accuracy by up to 16.4% on well-established text-to-code generation benchmarks. LPW also sets new state-of-the-art Pass@1 accuracy, achieving 98.2% on HumanEval, 84.8% on MBPP, 59.3% on LiveCode, 62.6% on APPS, and 34.7% on CodeContests, using GPT-4o as the backbone. Our code is publicly available at: https://github.com/you68681/lpw.

Planning-Driven Programming: A Large Language Model Programming Workflow

Large language models (LLMs) with one or more fine-tuning phases have become necessary to unlock various capabilities, enabling LLMs to follow natural language instructions and align with human preferences. However, it carries the risk of catastrophic forgetting during sequential training, the parametric knowledge or the ability learned in previous stages may be overwhelmed by incoming training data. This paper finds that LLMs can restore some original knowledge by regularly resetting partial parameters. Inspired by this, we introduce \underline{H}alf \underline{F}ine-\underline{T}uning (HFT) for LLMs, as a substitute for full fine-tuning (FFT), to mitigate the forgetting issues, where half of the parameters are selected to learn new tasks. In contrast, the other half are frozen to retain previous knowledge. We provide a feasibility analysis from the optimization perspective and interpret the parameter selection operation as a regularization term. HFT could be seamlessly integrated into existing fine-tuning frameworks without changing the model architecture. Extensive experiments and analysis on supervised fine-tuning, direct preference optimization, and continual learning consistently demonstrate the effectiveness, robustness, and efficiency of HFT. Compared with FFT, HFT not only significantly alleviates the forgetting problem, but also achieves the best performance in a series of downstream benchmarks, with an approximately 30% reduction in training time.

HFT: Half Fine-Tuning for Large Language Models

Commonsense question answering (QA) are widely used to evaluate the commonsense abilities of large language models. However, answering commonsense questions correctly requires not only knowledge but also reasoning—even for seemingly simple questions. We demonstrate that such hidden reasoning attributes in commonsense questions can lead evaluation accuracy differences of up to 24.8% across different difficulty levels in the same benchmark. Current benchmarks overlook these hidden reasoning attributes, making it difficult to assess a model's specific levels of commonsense knowledge and reasoning ability. To address this issue, we introduce ReComSBench, a novel framework that reveals hidden reasoning attributes behind commonsense questions by leveraging the knowledge generated during the reasoning process. Additionally, ReComSBench proposes three new metrics for decoupled evaluation: Knowledge Balanced Accuracy, Marginal Sampling Gain, and Knowledge Coverage Ratio. Experiments show that ReComSBench provides insights into model performance that traditional benchmarks cannot offer. The difficulty stratification based on revealed hidden reasoning attributes performs as effectively as the model-probability-based approach but is more generalizable and better suited for improving a model's commonsense reasoning abilities. By uncovering and analyzing the hidden reasoning attributes in commonsense data, ReComSBench offers a new approach to enhancing existing commonsense benchmarks.

Beyond Surface Simplicity: Revealing Hidden Reasoning Attributes for Precise Commonsense Diagnosis

Mixture-of-Experts (MoE) shines brightly in large language models (LLMs) and demonstrates outstanding performance in plentiful natural language processing tasks. However, existing methods transforming LLMs from dense to MoE face significant data requirements and typically rely on large-scale post-training. In this paper, we propose Upcycling Instruction Tuning (UpIT), a data-efficient approach for tuning a dense pre-trained model into a MoE instruction model. Specifically, we first point out that intermediate checkpoints during instruction tuning of the dense model are naturally suitable for specialized experts, and then propose an expert expansion stage to flexibly achieve models with flexible numbers of experts, where genetic algorithm and parameter merging are introduced to ensure sufficient diversity of new extended experts. To ensure that each specialized expert in the MoE model works as expected, we select a small amount of seed data that each expert excels to pre-optimize the router. Extensive experiments with various data scales and upcycling settings demonstrate the outstanding performance and data efficiency of UpIT, as well as stable improvement in expert or data scaling. Further analysis reveals the importance of ensuring expert diversity in upcycling.

Downloads

Next from ACL 2025

VMLU Benchmarks: A comprehensive benchmark toolkit for Vietnamese LLMs

Similar lecture

DAC: Decomposed Automation Correction for Text-to-SQL