Austria

ARC Challenge appears more difficult than ARC Easy for modern LLMs primarily due to an evaluation setup that prevents direct comparison of answer choices rather than inherent complexity. Although some researchers have quietly shifted to a more appropriate scheme over the last year, the implications of this change have yet to be widely acknowledged. We highlight this overlooked shift, show how similar evaluation practices falsely imply reasoning deficits in other benchmarks, and demonstrate that fairer methods dramatically reduce performance gaps (e.g. on SIQA) and even yield superhuman results (OpenBookQA). In doing so, we reveal how evaluation shapes perceived difficulty and offer guidelines to ensure that multiple-choice evaluations accurately reflect actual model capabilities.

ACL 2025

ARC &#39;Challenge&#39; Is Not That Challenging

multi-choice problems

evaluation

### Welcome to CogSci Conference 2025!

The 47th Annual Meeting of the Cognitive Science Society was a hybrid meeting held in San Francisco. 

<div style="position:relative;padding-top:0;width:900px;height:500px;"><iframe style="position:absolute;border:none;width:100%;height:100%;left:0;top:0;" src="https://online.fliphtml5.com/ebtyf/amvr/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe></div>

#### About

The Cognitive Science Society brings together researchers from around the world who hold a common goal: understanding the nature of the human mind. The mission of the Society is to promote Cognitive Science as a discipline, and to foster scientific interchange among researchers in various areas of study, including Artificial Intelligence, Linguistics, Anthropology, Psychology, Neuroscience, Philosophy, and Education.

The Society is a non-profit professional organization and its activities include sponsoring an annual conference and publishing the journals Cognitive Science and TopiCS.

#### Our History 

* **Society Creation**<br>
The Society was incorporated as a 501(c)(3) non-profit professional organization in Massachusetts in 1979. The organizing committee included Roger Schank, Allan Collins, Donald Norman, and a number of other scholars from psychology, linguistics, computer science, and philosophy. 
<br><br>
* **Conference Creation**<br>
The first conference on cognitive science was held at La Jolla, California in August, 1979, and has occurred annually since then. The proceedings of each conference are published, and those from most years are available through Lawrence Erlbaum Associates, Inc. The annual proceedings of the Cognitive Science Conference represent a major source of information on new work and new ideas in the scientific study of thinking. In 1990, the Society, with help from an anonymous donor, established the David Marr Prize for the best student paper at each annual meeting.
<br><br>
* **Journal Creation**<br>
The Journal, Cognitive Science, began publication in 1976, and is now published by Wiley-Blackwell. The Executive Editor is currently Richard P. Cooper of Birkbeck, University of London, and there are 18 Associate Editors and a 30-member editorial board. It serves as the premier outlet for research reports that intersect two or more disciplines. Copyrights for articles published in the journal are held by the Society. The Governing Board of the Cognitive Science Society voted in late 2006 to found a new journal, Topics in Cognitive Science (topiCS). The Editor in Chief is Wayne Gray, Cognitive Science Department, Rensselaer Polytechnic Institute. The journal seeks to fill a niche not occupied by Cognitive Science Journal or other cognitive science journals. Membership in the Society includes a subscription to Cognitive Science and TopiCS. Copyrights for articles published in the journal are held by the Society.
<br><br>

#### Code of Conduct

By attending the CogSci 2025 Conference, you are required to adhere to the society’s **[Code of Conduct](https://drive.google.com/file/d/1ChPuihLy6jE_BWqfO7J2KKgX35JW2zsM/view?usp=sharing)**.
<br><br>


You need to log in with the email address you registered with. 

Login credentials were sent to you from Underline -  subject line "Welcome to the CogSci 2025 Conference". Please be sure to check your spam/promotional inbox  if you do not see an email confirmation right away.





Please log in to join this event.

To access the site, please register [**here**](https://cognitivesciencesociety.org/registration/).

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

CogSci 2025

The 47th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

ARC 'Challenge' Is Not That Challenging

poster

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair:<br>
*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.*<br>
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)<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 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)

Recent generative large language models (LLMs) show remarkable performance in non-English languages, but when prompted in those languages they tend to express higher harmful social biases and toxicity levels. Prior work has shown that finetuning on specialized datasets can mitigate this behavior, and doing so in English can transfer to other languages. In this work, we investigate the impact of different finetuning methods on the model's bias and toxicity, but also on its ability to produce fluent and diverse text. We reduce biases by finetuning on curated non-harmful text, but find only direct preference optimization to be effective for mitigating toxicity. The mitigation caused by applying these methods in English also transfers to non-English languages. We find evidence that the extent to which transfer takes place can be predicted by the amount of data in a given language present in the model's pretraining data. However, this transfer of bias and toxicity mitigation often comes at the expense of decreased language generation ability in non-English languages, highlighting the importance of developing language-specific bias and toxicity mitigation methods.

Cross-Lingual Transfer of Debiasing and Detoxification in Multilingual LLMs: An Extensive Investigation

Motivated by the surge of large language models, there has been a push to formally characterize the symbolic abilities intrinsic to the transformer architecture. A programming language, called RASP, has been proposed, which can be directly compiled into transformer weights to implement these algorithms. However, the tasks that can be implemented in RASP are often uncommon to learn from natural unsupervised data, showing a mismatch between theoretical capabilities of the transformer architecture, and the practical learnability of these capabilities from unsupervised data. We propose tracr-injection, a method that allows us to distill algorithms written in RASP directly into a pre-trained language model. We showcase our method by injecting 3 different algorithms into a language model. We show how our method creates an interpretable subspace within the model's residual stream, which can be decoded into the variables present in the code of the RASP algorithm. Additionally, we found that the proposed method can improve out-of-distribution performance compared to our baseline, indicating that indeed a more symbolic mechanism is taking place in the inner workings of the model. We release the code used to run our experiments.

Tracr-Injection: Distilling Algorithms into Pre-trained Language Models

Optimizing Large Language Model (LLM) performance requires well-crafted prompts, but manual prompt engineering is labor-intensive and often ineffective. Automated prompt optimization techniques address this challenge but the major of them rely on randomly selected evaluation subsets, which fail to represent the full dataset, leading to unreliable evaluations and suboptimal prompts. Existing coreset selection methods, designed for LLM benchmarking, are unsuitable for prompt optimization due to challenges in clustering similar samples, high data collection costs, and the unavailability of performance data for new or private datasets. To overcome these issues, we propose IPOMP, an Iterative evaluation data selection approach for effective Prompt Optimization using real time Model Performance. IPOMP is a two-stage approach that selects representative and diverse samples using semantic clustering and boundary analysis, followed by iterative refinement with real-time model performance data to replace redundant samples. Evaluations on two datasets BIG-bench and LIAR, and two models GPT-3.5 and GPT-4o-mini, show that IPOMP improves effectiveness by at least 1.6% to 3.1%, and stability by at least 50% to 55.5% compared with the best baseline across the studied datasets and models, with minimal computational overhead below 1%. Furthermore, the results demonstrate that our real-time performance-guided refinement approach can be universally applied to enhance existing coreset selection methods.

Model Performance-Guided Evaluation Data Selection for Effective Prompt Optimization

As large language models advance toward superhuman performance, ensuring their alignment with human values and abilities grows increasingly complex. Weak-to-strong generalization offers a promising approach by leveraging predictions from weaker models to guide stronger systems, but its effectiveness could be constrained by the inherent noise and inaccuracies in these weak predictions. To address this, we propose a theoretically grounded approach that replaces forward KL divergence—whose mass-covering behavior risks overfitting to imperfect weak signals—with reverse KL divergence. Reverse KL divergence’s zero-forcing effect prioritizes high-confidence predictions, effectively mitigating the influence of unreliable weak supervision. Theoretically, we extend existing bounds and derive tighter lower bounds for both forward and reverse KL divergence. Notably, when a sufficiently pre-trained strong model is fine-tuned on the last linear layer, reverse KL guarantees that it outperforms its weak supervisor by the magnitude of their disagreement. Empirically, we demonstrate that reverse KL and reverse cross-entropy not only enable strong models to outperform those trained with forward KL and standard cross-entropy across most settings, but also exhibit greater robustness to noisy labels.

Revisiting Weak-to-Strong Generalization in Theory and Practice: Reverse KL vs. Forward KL

Due to the large number of parameters, the inference phase of Large Language Models (LLMs) is resource-intensive. Unlike traditional model compression, which needs retraining, recent dynamic computation methods show that not all components are required for inference, enabling a training-free pipeline.In this paper, we focus on the dynamic depth of LLM generation. A token-position aware layer skipping framework is proposed to save 1.5x times operations efficiently while maintaining performance.We first observed that tokens predicted later have lower perplexity and thus require less computation. Then, we propose a training-free algorithm called Position-Aware Depth Decay Decoding (), which leverages a power-law decay function, \left\lfloor L × (𝛼i) \right\rfloor, to determine the number of layers to retain when generating token Ti. Remarkably, without any retraining, the achieves success across a wide range of generation tasks for the first time.Experiments on large language models (the Llama) with 7 ∼ 70 billion parameters show that can achieve an average 1.5x speedup compared with the full-inference pipeline while maintaining comparable performance with nearly no performance drop (<1%) on the GSM8K and BBH benchmarks.

Position-Aware Depth Decay Decoding (D3): Boosting Large Language Model Inference Efficiency

Recent advancements in Generative AI and Large Language Models (LLMs) have enabled the creation of highly realistic synthetic content, raising concerns about the potential for malicious use, such as misinformation and manipulation. Moreover, detecting Machine-Generated Text (MGT) remains challenging due to the lack of robust benchmarks that assess generalization to real-world scenarios. In this work, we evaluate the resilience of state-of-the-art MGT detectors (e.g., Mage, Radar, LLM-DetectAIve) to linguistically informed adversarial attacks. We develop a pipeline that fine-tunes language models using Direct Preference Optimization (DPO) to shift the MGT style toward human-written text (HWT), obtaining generations more challenging to detect by current models. Additionally, we analyze the linguistic shifts induced by the alignment and how detectors rely on "linguistic shortcuts" to detect texts. Our results show that detectors can be easily fooled with relatively few examples, resulting in a significant drop in detecting performances. This highlights the importance of improving detection methods and making them robust to unseen in-domain texts. We release code, models, and data to support future research on more robust MGT detection benchmarks.

Stress-testing Machine Generated Text Detection: Shifting Language Models Writing Style to Fool Detectors

Simultaneous translation of unbounded streaming speech remains a challenging problem due to the need for effectively processing the history speech context and past translations so that quality and latency, including computation overhead, can be balanced. Most prior works assume pre-segmented speech, limiting their real-world applicability. In this paper, we propose InfiniSST, a novel approach that formulates SST as a multi-turn dialogue task, enabling seamless translation of unbounded speech. We construct translation trajectories and robust segments from MuST-C with multi-latency augmentation during training and develop a key-value (KV) cache management strategy to facilitate efficient inference. Experiments on MuST-C En-Es, En-De, and En-Zh demonstrate that InfiniSST reduces computation-aware latency by 0.5 to 1 second while maintaining the same translation quality compared to baselines. Ablation studies further validate the contributions of our data construction and cache management strategy.

InfiniSST: Simultaneous Translation of Unbounded Speech with Large Language Model

The rapid advancement of vision-language models (VLMs) has brought a lot of attention to their safety alignment. However, existing methods have primarily focused on model undersafety, where the model responds to hazardous queries, while neglecting oversafety, where the model refuses to answer safe queries. In this paper, we introduce the concept of safety calibration, which systematically addresses both undersafety and oversafety. Specifically, we present VSCBench, a novel dataset of 3,600 image-text pairs that are visually or textually similar but differ in terms of safety, which is designed to evaluate safety calibration across image-centric and text-centric scenarios. Based on our benchmark, we evaluate safety calibration across eleven widely used VLMs. Our extensive experiments revealed major issues with both undersafety and oversafety. We further investigated four approaches to improve the model's safety calibration. We found that even though some methods effectively calibrated the models' safety problems, these methods also lead to the degradation of models' utility. This trade-off underscores the urgent need for advanced calibration methods, and our benchmark provides a valuable tool for evaluating future approaches.

VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration

Recent advances in mathematical problem-solving with language models (LMs) integrate chain-of-thought (CoT) reasoning and code execution to harness their complementary strengths. However, existing hybrid frameworks exhibit a critical limitation: they depend on externally dictated instructions or rigid code-integration templates, lacking metacognitive awareness—the capacity to dynamically evaluate intrinsic capabilities and autonomously determine when and how to integrate tools. This rigidity motivates our study of autonomous code integration, enabling models to adapt tool-usage strategies as their reasoning abilities evolve during training.

While reinforcement learning (RL) shows promise for boosting LLM reasoning at scale (e.g., DeepSeek-R1), we demonstrate its inefficiency in learning autonomous code integration due to inadequate exploration of the vast combinatorial space of CoT-code interleaving patterns. To address this challenge, we propose a novel Expectation-Maximization (EM) framework that synergizes structured exploration (E-step) with off-policy RL optimization (M-step), creating a self-reinforcing cycle between metacognitive tool-use decisions and evolving capabilities. Experiments reveal our method achieves superior results through improved exploration. Notably, our 7B model improves over 11% on MATH500 and 9.4% on AIME without o1-like CoT.

Learning Autnomous Code Integration for Math Language Models

Large Language Models (LLMs) often succumb to adversarial prompts, a phenomenon popularly known as ``jailbreaking.'' While jailbreaking primarily targets short-term noncompliance with predefined policies, we argue that a deeper vulnerability lies in altering an LLM's \emph{fundamental axiomatic beliefs}, such as mathematical or philosophical truths. In this work, we introduce \textsc{GoodLiar}, a reinforcement learning (RL)-based framework that generates deceptive contexts to systematically \emph{rewrite} an LLM's core logical or philosophical understandings. By incentivizing an RL agent to produce persuasive and coherent arguments, \textsc{GoodLiar} aims to induce \emph{persistent} belief shifts, rather than merely influencing immediate judgments of factual truthfulness. %rather than one-off policy breaches. Our approach introduces \emph{DA-ILQL}, a novel offline RL method that extends ILQL by integrating on-policy data and language exploration to enhance the language discovery and optimization. Through extensive evaluations on multiple LLMs, we show that deceptive contexts discovered by \textsc{GoodLiar} consistently outperform simple multi-turn prompting methods.

Downloads

Next from ACL 2025

Cross-Lingual Transfer of Debiasing and Detoxification in Multilingual LLMs: An Extensive Investigation

Similar lecture

A Neurosymbolic Model of Human Reasoning on the Abstraction and Reasoning Corpus

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES