On Representational Dissociation of Language and Arithmetic in Large Language Models

The association between language and (non-linguistic) thinking ability in humans has long been debated, and recently, neuroscientific evidence of brain activity patterns has been considered. Such a scientific context naturally raises an interdisciplinary question -- what about such a language-thought dissociation in large language models (LLMs)? In this paper, as an initial foray, we explore this question by focusing on simple arithmetic skills (e.g., $1+2=$ ?) as a thinking ability and analyzing the geometry of their encoding in LLMs' representation space. Our experiments with linear classifiers and cluster separability tests demonstrate that simple arithmetic equations and general language input are encoded in completely separated regions in LLMs' internal representation space across all the layers, which is also supported with more controlled stimuli (e.g., spelled-out equations). These tentatively suggest that arithmetic reasoning is mapped into a distinct region from general language input, which is in line with the neuroscientific observations of human brain activations, while we also point out their somewhat cognitively implausible geometric properties.

Can Language Models Learn Typologically Implausible Languages?

Grammatical features across human languages show intriguing correlations often attributed to learning biases in humans. However, empirical evidence has been limited to experiments with highly simplified artificial languages, and whether these correlations arise from domain-general or language-specific biases remains a matter of debate. Language models (LMs) provide an opportunity to study artificial language learning at a large scale and with a high degree of naturalism. In this paper, we begin with an in-depth discussion of how LMs allow us to better determine the role of domain-general learning biases in language universals. We then assess learnability differences for LMs resulting from typologically plausible and implausible languages closely following the word-order universals identified by linguistic typologists. We conduct a symmetrical cross-lingual study training and testing LMs on an array of highly naturalistic but counterfactual versions of the English (head-initial) and Japanese (head-final) languages. Compared to similar work, our datasets are more naturalistic and fall closer to the boundary of plausibility. Our experiments show that these LMs are often slower to learn these subtly implausible languages, while ultimately achieving similar performance on some metrics regardless of typological plausibility. These findings lend credence to the conclusion that LMs do show some typologically-aligned learning preferences, and that the typological patterns may result from, at least to some degree, domain-general learning biases.

Large Language Models Are Human-Like Internally

Recent cognitive modeling studies have reported that larger language models (LMs) exhibit a poorer fit to human reading behavior, leading to claims of their cognitive implausibility. In this paper, we revisit this argument through the lens of mechanistic interpretability and argue that prior conclusions were skewed by an exclusive focus on the final layers of LMs. Our analysis reveals that next-word probabilities derived from internal layers of larger LMs align with human sentence processing data as well as, or better than, those from smaller LMs. This alignment holds consistently across behavioral (self-paced reading times, gaze durations, MAZE task processing times) and neurophysiological (N400 brain potentials) measures, challenging earlier mixed results and suggesting that the cognitive plausibility of larger LMs has been underestimated. Furthermore, we first identify an intriguing relationship between LM layers and human measures: earlier layers correspond more closely with fast gaze durations, while later layers better align with relatively slower signals such as N400 potentials and MAZE processing times. Our work opens new avenues for interdisciplinary research at the intersection of mechanistic interpretability and cognitive modeling.

Libra-Leaderboard: Towards Responsible AI through a Balanced Leaderboard of Safety and Capability

To address this gap, we introduce Libra-Leaderboard, a comprehensive framework designed to rank LLMs through a balanced evaluation of performance and safety. Combining a dynamic leaderboard with an interactive LLM arena, Libra-Leaderboard encourages the joint optimization of capability and safety. Unlike traditional approaches that average performance and safety metrics, Libra-Leaderboard uses a distance-to-optimal-score method to calculate the overall rankings. This approach incentivizes models to achieve a balance rather than excelling in one dimension at the expense of some other ones. In the first release, Libra-Leaderboard evaluates 26 mainstream LLMs from 14 leading organizations, identifying critical safety challenges even in state-of-the-art models.

Can Input Attributions Interpret the Inductive Reasoning Process Elicited in In-Context Learning?

Elucidating the rationale behind neural models' outputs has been challenging in the machine learning field, which is indeed applicable in this age of large language models (LLMs) and in-context learning (ICL). When it comes to estimating input attributions (IA), ICL poses a new issue of interpreting which example in the prompt, consisting of a set of examples, contributed to identifying the task/rule to be solved. To this end, in this paper, we introduce synthetic diagnostic tasks inspired by the poverty of the stimulus design in inductive reasoning; here, most in-context examples are ambiguous w.r.t. their underlying rule, and one critical example disambiguates the task demonstrated. The question is whether conventional IA methods can identify such an example in interpreting the inductive reasoning process in ICL. Our experiments provide several practical findings; for example, a certain simple IA method works the best, and the larger the model, the generally harder it is to interpret the ICL with gradient-based IA methods.

Think-to-Talk or Talk-to-Think? When LLMs Come Up with an Answer in Multi-Step Reasoning

This study investigates the internal reasoning mechanism of language models during symbolic multi-step reasoning, motivated by the question of whether chain-of-thought (CoT) outputs are faithful to the model's internals. Specifically, we inspect when they internally determine their answers, particularly before or after CoT begins, to determine whether models follow a post-hoc "think-to-talk" mode or a step-by-step "talk-to-think" mode of explanation. Through causal probing experiments in controlled arithmetic reasoning tasks, we found systematic internal reasoning patterns across models; for example, simple subproblems are solved before CoT begins, and more complicated multi-hop calculations are performed during CoT.

First Heuristic Then Rational: Dynamic Use of Heuristics in Language Model Reasoning

Multi-step reasoning is widely adopted in the community to explore the better performance of language models (LMs). We report on the systematic strategy that LMs use in this process. Our controlled experiments reveal that LMs rely more heavily on heuristics, such as lexical overlap, in the earlier stages of reasoning when more steps are required to reach an answer. Conversely, as LMs progress closer to the final answer, their reliance on heuristics decreases. This suggests that LMs track only a limited number of future steps and dynamically combine heuristic strategies with logical ones in tasks involving multi-step reasoning.

CVQA: Culturally-diverse Multilingual Visual Question Answering Benchmark

Visual Question Answering (VQA) is an important task in multimodal AI, and it is often used to test the ability of vision-language models to understand and reason on knowledge present in both visual and textual data. However, most of the current VQA models use datasets that are primarily focused on English and a few major world languages, with images that are typically Western-centric. While recent efforts have tried to increase the number of languages covered on VQA datasets, they still lack diversity in low-resource languages. More importantly, although these datasets often extend their linguistic range via translation or some other approaches, they usually keep images the same, resulting in narrow cultural representation. To address these limitations, we construct CVQA, a new Culturally-diverse multilingual Visual Question Answering benchmark, designed to cover a rich set of languages and cultures, where we engage native speakers and cultural experts in the data collection process. As a result, CVQA includes culturally-driven images and questions from across 28 countries on four continents, covering 26 languages with 11 scripts, providing a total of 9k questions. We then benchmark several Multimodal Large Language Models (MLLMs) on CVQA, and show that the dataset is challenging for the current state-of-the-art models. This benchmark can serve as a probing evaluation suite for assessing the cultural capability and bias of multimodal models and hopefully encourage more research efforts toward increasing cultural awareness and linguistic diversity in this field.

Emergent Word Order Universals from Cognitively-Motivated Language Models

The world's languages exhibit certain so-called typological or implicational universals; for example, Subject-Object-Verb (SOV) word order typically employs postpositions. Explaining the source of such biases is a key goal in linguistics. We study the word-order universals through a computational simulation with language models (LMs). Our experiments show that typologically typical word orders tend to have lower perplexity estimated by LMs with cognitively plausible biases: syntactic biases, specific parsing strategies, and memory limitations. This suggests that the interplay of these cognitive biases and predictability (perplexity) can explain many aspects of word-order universals. This also showcases the advantage of cognitively-motivated LMs, which are typically employed in cognitive modeling, in the computational simulation of language universals.

Psychometric Predictive Power of Large Language Models

Next-word probabilities from language models have been shown to successfully simulate human reading behavior. Building on this, we show that, interestingly, instruction-tuned large language models (LLMs) yield worse psychometric predictive power (PPP) for human reading behavior than base LLMs with equivalent perplexities. In other words, instruction tuning, which helps LLMs provide human-preferred responses, does not always make them human-like from the computational psycholinguistics perspective. In addition, we explore prompting methodologies in simulating human reading behavior with LLMs, showing that prompts reflecting a particular linguistic hypothesis lead LLMs to exhibit better PPP but are still worse than base LLMs. These highlight that recent instruction tuning and prompting do not offer better estimates than direct probability measurements from base LLMs in cognitive modeling.