United States

Reliable automatic evaluation of retrieval-grounded long-form reports typically requires human annotation or frontier-scale proprietary LLMs, both of which are expensive in constrained settings. Team rgipt participated in RAG4Reports@ACL 2026 Task 1 with a zero-shot nugget-verification system that runs entirely on a single NVIDIA T4 GPU. We compare three ultra-lightweight decoder-only models: Qwen2-0.5B, Qwen2-1.5B, and Qwen2.5-0.5B, under identical inference conditions to examine how small an LLM judge can be while retaining human-aligned ranking signal. Both Qwen2 models produced negative $\tau_{\text{gap}}$, whereas Qwen2.5-0.5B achieved $\tau_{\text{gap}} = 0.0772$ and Pearson $r = 0.2209$, ranking 13th of 21 teams. Within this family and evaluation setting, model generation appears to matter more than parameter count, although this finding is based on three configurations on a single task and warrants further validation.

ACL 2026

Exploring Capability Thresholds in Ultra-Lightweight LLM Judges for Nugget-Based Report Evaluation

automatic report evaluation

ultra-lightweight language models

llm-as-a-judge

### 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**. 

ACL 2025

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

workshop 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.

**Online Registration Form**: https://acl.swoogo.com/acl2026

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.

We describe EFSG (Evidence-First Structured Generation), our submission to Task B of the RAG4Reports@ACL 2026 shared task. Standard retrieval-augmented generation pipelines allow generation models to write from parametric memory and attach citations retroactively: a behaviour we term post-rationalization. EFSG addresses this structurally through a phase boundary: all evidence is retrieved, extracted, and sealed into a fact pool before any generation begins; each sentence then sees only its single committed source passage. Our best run (t5100k doc corpus) achieved sentence_support of 0.612 and nugget_coverage of 0.126 (F1 = 0.182).

EFSG: Evidence-First Structured Generation for Multilingual RAG Report Generation

In the present article, we have described our system developed for participating in Task B on Multilingual Report Generation under RAG4Reports 2026 at ACL 2026 with submitted run ID ju_nlp_pg. The problem statement is given a report request in English, the system retrieves relevant passages from a four million multilingual document corpus (English, Chinese, Russian, Arabic) and generates a grounded, citation-bearing report. Our core challenge was how to fit a large retrieval corpus along with a capable generative model on a two-GPU node with ‚âà29 GB RAM. We addressed the challenge employing three different techniques: (1) 4-bit NF4 quantization, shrinking the LLM from ‚âà14 GB to ‚âà4 GB; (2) memory-mapped, chunked FAISS index construction over pre-computed multilingual-e5-large embeddings; and (3) strict model-loading order to prevent heap fragmentation. On the other hand, the reports are structured around topic nuggets to directly target the Auto-ARGUE evaluation signal.

JU-NLP-PG at RAG4Reports 2026: Memory-Efficient Multilingual Report Generation with 4-bit Quantized LLMs

Quantization methods are widely used to accelerate inference and streamline the deployment of large language models (LLMs). Although quantization's effects on various LLM capabilities have been extensively studied, one critical area remains underexplored: factual knowledge recall (FKR), the process by which LLMs access stored knowledge. To this end, we conduct comprehensive experiments using three common quantization techniques at distinct bit widths, in conjunction with interpretability-driven analyses on two tasks, knowledge memorization and latent multi-hop reasoning. We show that quantization typically results in information loss within LLMs, consequently diminishing their capacity for FKR. This effect is particularly amplified in smaller models within the same architectural families. However, models quantized at reduced bit precision do not consistently exhibit inferior performance and occasionally quantization may even enhance model FKR. We find that BitSandBytes demonstrates highest preservation of the original full-precision model's FKR. Despite variability across models and methods, quantization causes modest performance degradation and remains an effective compression strategy.

Through a Compressed Lens: Investigating The Impact of Quantization on Factual Knowledge Recall

Recent research has increasingly focused on training large language models (LLMs) using federated learning, known as FedLLM. However, responsible AI (RAI)—which aims to ensure safe and trustworthy responses—remains underexplored in this context. In FedLLM, client-side training data may contain harmful content, resulting in unsafe LLMs that can generate inappropriate responses. Aggregating such models into a global model and redistributing it to clients risks the widespread deployment of unsafe LLMs. To address this, we incorporate two well-established RAI techniques into FedLLM: safety filtering and constitutional AI. Our experiments show that these methods significantly improve LLM safety, achieving over 20% improvement on AdvBench.

Responsible Federated LLMs via Safety Filtering and Constitutional AI

Detecting media bias is increasingly critical in today’s fast-evolving digital landscape. While large language models (LLM) show promise in bias detection, their reasoning is often unreliable, producing opaque errors. To address this, we propose an {\bf uncertainty-aware proxy reasoning method} that combines proxy attribute prediction with uncertainty estimation to assess and stratify reasoning quality. By modeling uncertainty, our method reliably narrows down the intermediate reasoning space and highlights trustworthy bias indicators, enhancing transparency and interpretability for end users and downstream applications. We evaluate our method on the BASIL benchmark and compare it against strong LLM baselines and recent state-of-the-art approaches. Our experiments assess both decision accuracy and quality. Results show a 6.25% relative F1 improvement over competitive baselines, demonstrating that uncertainty-aware reasoning boosts performance and provides insight into the decision-making process, making media bias detection more reliable and interpretable.

Uncertainty-Aware Proxy Attribute Reasoning for Reliable Media Bias Detection

Large language models (LLMs) are increasingly deployed in wide range of applications, yet remain vulnerable to adversarial jailbreak attacks that circumvent their safety guardrails.
Existing evaluation frameworks typically report binary success/failure metrics, failing to capture the temporal dynamics of how attacks succeed under persistent adversarial pressure. This preliminary work proposes a novel evaluation framework that applies survival analysis techniques to characterize LLM jailbreak vulnerability. Our approach models the “time-to-jailbreak” as a survival outcome, enabling estimation of hazard functions, survival curves, and risk factors associated with successful attacks. We evaluate three LLMs against a sub-set of prompts from the HarmBench dataset spanning three attack categories. Our analysis reveals that models exhibit distinct vulnerability profiles: while one model demonstrates rapid degradation under iterative attacks, the wo other models show consistent moderate vulnerability. Our framework provides actionable insights for model and LLM applicaiton developers and establishes survival analysis as a rigorous methodology for LLM safety evaluation.

Quantifying LLM Safety Degradation Under Repeated Attacks Using Survival Analysis

Verifying complex real-world claims against diverse and potentially unreliable open-web sources requires balancing evidence comprehensiveness with rigorous source reliability. Current automated fact-checking approaches often fail to address this holistically, losing contextual dependencies and applying trust signals monolithically at the document level.

We introduce ClaimCLAIRE, a multi-component fact-checking agent that integrates four key innovations: (1) iterative component-aware decomposition with exhaustiveness validation, (2) holistic evidence gathering using a ReAct agent that maintains cross-component semantic awareness, (3) trust-modulated retrieval that weights evidence by source credibility to mitigate the influence of misinformation, and (4) adaptive gap-filling to address recall bottlenecks in under-supported sub-claims.

Evaluated on the AVeriTeC benchmark, ClaimCLAIRE achieves 84.27% accuracy and a macro-F1 of 0.806. Our systematic ablations demonstrate that while decomposition alone can degrade performance, its integration with trust-aware retrieval and adaptive gap-filling yields a pipeline where component-level verdicts, source trust ratings, and deterministic AND-logic synthesis together support transparent, accountable fact verification.

ClaimCLAIRE: A Trust-Aware Multi-Component Fact-Checking Agent for Open-World Claims

This paper introduces ChatbotManip, a novel dataset for studying manipulation in Chatbots. It contains simulated generated conversations between a chatbot and a (simulated) user, where the chatbot is explicitly asked to showcase manipulation tactics, persuade the user towards some goal, or simply be helpful. We consider a diverse set of chatbot manipulation contexts, from consumer and personal advice to citizen advice and controversial proposition argumentation. Each conversation is annotated by human annotators for both general manipulation and specific manipulation tactics. Our research reveals three key findings. First, Large Language Models (LLMs) can be manipulative when explicitly instructed, with annotators identifying manipulation in approximately 84% of such conversations. Second, even when only instructed to be ``persuasive'' without explicit manipulation prompts, LLMs frequently default to controversial manipulative strategies, particularly Gaslighting and Fear Enhancement. Third, zero-shot larger models such as Gemini 2.5 pro have the best performance in detecting manipulation (of the models tested), with more work required to fine-tune smaller open source models for real-world on-device oversight. Our work provides important insights for AI safety research and highlights the need of addressing manipulation risks as LLMs are increasingly deployed in consumer-facing applications.

ChatbotManip: a Dataset to Facilitate Evaluation and Oversight of Manipulative Chatbot Behaviour

The fairness-accuracy trade-off is a key challenge in NLP tasks. Current work focuses on finding a single optimal solution to balance the two objectives, which is limited considering the diverse solutions on the Pareto front.
This work intends to provide controllable trade-offs according to the user's preference of the two objectives, which is defined as a reference vector. 
To achieve this goal, we apply multi-objective optimization (MOO), which can find solutions from various regions of the Pareto front. 
However, it is challenging to precisely control the trade-off due to the stochasticity of the training process and the high dimensional gradient vectors.
Thus, we propose Controllable Pareto Trade-off (CPT) that can effectively train models to perform different trade-offs according to users' preferences.
CPT 1) stabilizes the fairness update with a moving average of stochastic gradients to determine the update direction, and 2) prunes the gradients by only keeping the gradients of the critical parameters. 
We evaluate CPT on hate speech detection and occupation classification tasks. Experiments show that CPT can achieve a higher-quality set of solutions on the Pareto front than the baseline methods. It also exhibits better controllability and can precisely follow the human-defined reference vectors.

Controllable Pareto Trade-off between Fairness and Accuracy

As the influence of LLMs expands, it is imperative to gain insight into their decisions. One way to do that is to develop probes that detect the presence or absence of a broad set of concepts within the embeddings computed in an LLM - which is what we might say a model is ``thinking" about. Such probes should be low-cost and easily applicable to any LLM, so that monitoring for many concepts is possible during normal operation.

In this paper, we take the first steps towards developing the capability of creating many such probes by defining and executing examples of the key tasks needed: first, the careful delineation of a concept through the creation of a dataset with the concept both present and then absent. Then, the training and testing of a set of linear probes to detect the concept on any layer of an LLM, including an exploration of the complexity of the probe needed. Finally, we show that such probes can track concepts across larger contexts. This is done with four separate concepts and three different LLMs. When this process is scaled to many more concepts, it will create the ability to easily monitor new models.

Downloads

Next from ACL 2026

EFSG: Evidence-First Structured Generation for Multilingual RAG Report Generation

Similar lecture

JuStRank: Benchmarking LLM Judges for System Ranking

.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

EFSG: Evidence-First Structured Generation for Multilingual RAG Report Generation

Similar lecture

JuStRank: Benchmarking LLM Judges for System Ranking

Downloads