Recent research in vision-language models (VLMs) has centered around the possibility of equipping them with implicit long-form chainof-thought reasoning-akin to the success observed in language models-via distillation and reinforcement learning.But what about the non-reasoning models already trained and deployed across the internet?Should we simply abandon them, or is there hope for a search mechanism that can elicit hidden knowledge and induce long reasoning traces-without any additional training or supervision?In this paper, we explore this possibility using a Monte Carlo Tree Search (MCTS)-inspired algorithm, which injects subquestion-subanswer pairs into the model's output stream.We show that framing reasoning as a search process-where subquestions act as latent decisions within a broader inference trajectory-helps the model "connect the dots" between fragmented knowledge and produce extended reasoning traces in nonreasoning models.We evaluate our method across three benchmarks and observe consistent improvements.Notably, our approach yields a 2% overall improvement on MMMU-PRO, including a significant 9% gain in Liberal Arts.Problem (P) What is the position of the blue car in the image?A: parked on the sidewalk B: driving on the road C: parked on the grass D: parked on the road CoT: To determine the position of the blue car in the image, let's analyze the scene: * The image shows a busy urban street with several vehicles.* The blue car is located on the right side of the image.* It is positioned on what appears to be a paved surface, which is consistent with a road.* The car is not on the grass or the sidewalk, as those areas are clearly distinguishable in the image.* The car is stationary, suggesting it is parked.Given these observations, the blue car is parked on the road.

Aly M. Kassem, Omar Mahmoud, Niloofar Mireshghallah, Hyunwoo Kim, Yulia Tsvetkov, Yejin Choi, Sherif Saad, Santu Rana. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

While humans naturally develop theory of mind (ToM), the capability to understand other people's mental states and beliefs, state-ofthe-art large language models (LLMs) underperform on simple ToM benchmarks.We posit that we can extend our understanding of LLMs' ToM abilities by evaluating key human ToM precursors-perception inference and perception-to-belief inference-in LLMs.We introduce two datasets, Percept-ToMi and Percept-FANToM, to evaluate these precursory inferences for ToM in LLMs by annotating characters' perceptions on ToMi and FANToM, respectively.Our evaluation of eight state-ofthe-art LLMs reveals that the models generally perform well in perception inference while exhibiting limited capability in perception-tobelief inference (e.g., lack of inhibitory control).Based on these results, we present PercepToM, a novel ToM method leveraging LLMs' strong perception inference capability while supplementing their limited perception-to-belief inference.Experimental results demonstrate that PercepToM significantly enhances LLM's performance, especially in false belief scenarios.

Recent advances in large language models (LLM) have enabled richer social simulations, allowing for the study of various social phenomena.However, most recent work has used a more omniscient perspective on these simulations (e.g., single LLM to generate all interlocutors), which is fundamentally at odds with the non-omniscient, information asymmetric interactions that involve humans and AI agents in the real world.To examine these differences, we develop an evaluation framework to simulate social interactions with LLMs in various settings (omniscient, non-omniscient).Our experiments show that LLMs perform better in unrealistic, omniscient simulation settings but struggle in ones that more accurately reflect real-world conditions with information asymmetry.Our findings indicate that addressing information asymmetry remains a fundamental challenge for LLM-based agents.

Existing efforts on quantifying privacy implications for large language models (LLMs) solely focus on measuring leakage of training data. In this work, we shed light on the often-overlooked interactive settings where an LLM receives information from multiple sources and generates an output to be shared with other entities, creating the potential of exposing sensitive input data in inappropriate contexts. In these scenarios, humans nat- urally uphold privacy by choosing whether or not to disclose information depending on the context. We ask the question “Can LLMs demonstrate an equivalent discernment and reasoning capability when considering privacy in context?” We propose CONFAIDE, a benchmark grounded in the theory of contextual integrity and designed to identify critical weaknesses in the privacy reasoning capabilities of instruction-tuned LLMs. CONFAIDE consists of four tiers, gradually increasing in complexity, with the final tier evaluating contextual privacy reasoning and theory of mind capabilities. Our experiments show that even commercial models such as GPT-4 and ChatGPT reveal private information in contexts that humans would not, 39% and 57% of the time, respectively, highlighting the urgent need for a new direction of privacy-preserving approaches as we demonstrate a larger underlying problem stemmed in the models’ lack of reasoning capabilities.

Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Bras, Malihe Alikhani, Gunhee Kim, Maarten Sap, Yejin Choi. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.

Theory of mind (ToM) evaluations currently focus on testing models using passive narratives that inherently lack interactivity.We introduce FANTOM, a new benchmark designed to stress-test ToM within information-asymmetric conversational contexts via question answering.Our benchmark draws upon important theoretical requisites from psychology and necessary empirical considerations when evaluating large language models (LLMs).In particular, we formulate multiple types of questions that demand the same underlying reasoning to identify illusory or false sense of ToM capabilities in LLMs.We show that FANTOM is challenging for state-of-the-art LLMs, which perform significantly worse than humans even with chainof-thought reasoning or fine-tuning.1 Linda: Yeah, I got a golden retriever.She's so adorable.David: What's her favorite food?Kailey: Hey guys, I'

Most existing dialogue systems fail to respond properly to potentially unsafe user utterances by either ignoring or passively agreeing with them.To address this issue, we introduce PROSOCIALDIALOG, the first large-scale multi-turn dialogue dataset to teach conversational agents to respond to problematic content following social norms.Covering diverse unethical, problematic, biased, and toxic situations, PROSOCIALDIALOG contains responses that encourage prosocial behavior, grounded in commonsense social rules (i.e., rules-ofthumb, RoTs).Created via a human-AI collaborative framework, PROSOCIALDIALOG consists of 58K dialogues, with 331K utterances, 160K unique RoTs, and 497K dialogue safety labels accompanied by free-form rationales.With this dataset, we introduce a dialogue safety detection module, Canary, capable of generating RoTs given conversational context, and a socially-informed dialogue agent, Prost.Empirical results show that Prost generates more socially acceptable dialogues compared to other state-of-the-art language and dialogue models in both in-domain and out-of-domain settings.Additionally, Canary effectively guides off-the-shelf language models to generate significantly more prosocial responses.Our work highlights the promise and importance of creating and steering conversational AI to be socially responsible.

Empathy is a complex cognitive ability based on the reasoning of others' affective states.In order to better understand others and express stronger empathy in dialogues, we argue that two issues must be tackled at the same time: (i) identifying which word is the cause for the other's emotion from his or her utterance and (ii) reflecting those specific words in the response generation.However, previous approaches for recognizing emotion cause words in text require sub-utterance level annotations, which can be demanding.Taking inspiration from social cognition, we leverage a generative estimator to infer emotion cause words from utterances with no word-level label.Also, we introduce a novel method based on pragmatics to make dialogue models focus on targeted words in the input during generation.Our method is applicable to any dialogue models with no additional training on the fly.We show our approach improves multiple best performing dialogue agents on generating more focused empathetic responses in terms of both automatic and human evaluation.

Byeongchang Kim, Hyunwoo Kim, Seokhee Hong, Gunhee Kim. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

We explore the task of improving persona consistency of dialogue agents.Recent models tackling consistency often train with additional Natural Language Inference (NLI) labels or attach trained extra modules to the generative agent for maintaining consistency.However, such additional labels and training can be demanding.Also, we find even the bestperforming persona-based agents are insensitive to contradictory words.Inspired by social cognition and pragmatics, we endow existing dialogue agents with public self-consciousness on the fly through an imaginary listener.Our approach, based on the Rational Speech Acts framework (Frank and Goodman, 2012), can enforce dialogue agents to refrain from uttering contradiction.We further extend the framework by learning the distractor selection, which has been usually done manually or randomly.Results on Dialogue NLI (Welleck et al., 2019) and PersonaChat (Zhang et al., 2018) dataset show that our approach reduces contradiction and improves consistency of existing dialogue models.Moreover, we show that it can be generalized to improve contextconsistency beyond persona in dialogues.

Exploration based on state novelty has brought great success in challenging reinforcement learning problems with sparse rewards. However, existing novelty-based strategies become inefficient in real-world problems where observation contains not only task-dependent state novelty of our interest but also task-irrelevant information that should be ignored. We introduce an information- theoretic exploration strategy named Curiosity-Bottleneck that distills task-relevant information from observation. Based on the information bottleneck principle, our exploration bonus is quantified as the compressiveness of observation with respect to the learned representation of a compressive value network. With extensive experiments on static image classification, grid-world and three hard-exploration Atari games, we show that Curiosity-Bottleneck learns an effective exploration strategy by robustly measuring the state novelty in distractive environments where state-of-the-art exploration methods often degenerate.