Language models trained via outcome-based reinforcement learning (RL) to reason using chain-of-thought (CoT) have shown remarkable performance. Monitoring such a model's CoT may allow us to understand its intentions and detect potential malicious behavior. However, to be effective, this requires that CoTs are legible and faithful. We study CoT legibility across 14 reasoning models, finding that RL often causes reasoning to become illegible to both humans and AI monitors, with reasoning models (except Claude) generating illegible CoTs while returning to perfectly readable final answers. We show that models use illegible reasoning to reach correct answers (accuracy dropping by 53\% when forced to use only legible portions), yet find no correlation between legibility and performance when resampling - suggesting the relationship is more nuanced. We also find that legibility degrades on harder questions. We discuss potential hypotheses for these results, including steganography, training artifacts, and vestigial tokens. These results suggest that without explicit optimization for legibility, outcome-based RL naturally produces models with increasingly opaque reasoning processes, potentially undermining monitoring approaches.

While Chain-of-Thought (CoT) monitoring offers a unique opportunity for AI safety, this opportunity could be lost through shifts in training practices or model architecture. To help preserve monitorability, we propose a pragmatic way to measure two components of it: legibility (whether the reasoning can be followed by a human) and coverage (whether the CoT contains all the reasoning needed for a human to also produce the final output). We implement these metrics with an autorater prompt that enables any capable LLM to compute the legibility and coverage of existing CoTs. After sanity-checking our prompted autorater with synthetic CoT degradations, we apply it to several frontier models on challenging benchmarks, finding that they exhibit high monitorability. We present these metrics, including our complete autorater prompt, as a tool for developers to track how design decisions impact monitorability. While the exact prompt we share is still a preliminary version under ongoing development, we are sharing it now in the hopes that others in the community will find it useful. Our method helps measure the default monitorability of CoT - it should be seen as a complement, not a replacement, for the adversarial stress-testing needed to test robustness against deliberately evasive models.

Legibility of robot motion is critical in human-robot interaction, as it allows humans to quickly infer a robot's intended goal. Although traditional trajectory generation methods typically prioritize efficiency, they often fail to make the robot's intentions clear to humans. Meanwhile, existing approaches to legible motion usually produce only a single "most legible" trajectory, overlooking the need to modulate intent expressiveness in different contexts. In this work, we propose a novel motion generation framework that enables controllable legibility across the full spectrum, from highly legible to highly ambiguous motions. We introduce a modeling approach based on an Information Potential Field to assign continuous legibility scores to trajectories, and build upon it with a two-stage diffusion framework that first generates paths at specified legibility levels and then translates them into executable robot actions. Experiments in both 2D and 3D reaching tasks demonstrate that our approach produces diverse and controllable motions with varying degrees of legibility, while achieving performance comparable to SOTA. Code and project page: https://legibility-modulator.github.io.

As machine learning systems increasingly inform critical decisions, the need for human-understandable explanations grows. Current evaluations of Explainable AI (XAI) often prioritize technical fidelity over cognitive accessibility which critically affects users, in particular those with visual impairments. We propose CUE, a model for Cognitive Understanding of Explanations, linking explanation properties to cognitive sub-processes: legibility (perception), readability (comprehension), and interpretability (interpretation). In a study (N=455) testing heatmaps with varying col-ormaps (BWR, Cividis, Coolwarm), we found comparable task performance but lower confidence/effort for visually impaired users. Unlike expected, these gaps were not mitigated and sometimes worsened by accessibility-focused color maps like Cividis. These results challenge assumptions about perceptual optimization and support the need for adaptive XAI interfaces. They also validate CUE by demonstrating that altering explanation legibility affects understandability. We contribute: (1) a formalized cognitive model for explanation understanding, (2) an integrated definition of human-centered explanation properties, and (3) empirical evidence motivating accessible, user-tailored XAI.

In this paper, we propose TrInk, a Transformer-based model for ink generation, which effectively captures global dependencies. To better facilitate the alignment between the input text and generated stroke points, we introduce scaled positional embeddings and a Gaussian memory mask in the cross-attention module. Additionally, we design both subjective and objective evaluation pipelines to comprehensively assess the legibility and style consistency of the generated handwriting. Experiments demonstrate that our Transformer-based model achieves a 35.56\% reduction in character error rate (CER) and an 29.66% reduction in word error rate (WER) on the IAM-OnDB dataset compared to previous methods. We provide an demo page with handwriting samples from TrInk and baseline models at: https://akahello-a11y.github.io/trink-demo/

Human--robot interaction requires robots whose actions are legible, allowing humans to interpret, predict, and feel safe around them. This study investigates the legibility of humanoid robot arm movements in a pointing task, aiming to understand how humans predict robot intentions from truncated movements and bodily cues. We designed an experiment using the NICO humanoid robot, where participants observed its arm movements towards targets on a touchscreen. Robot cues varied across conditions: gaze, pointing, and pointing with congruent or incongruent gaze. Arm trajectories were stopped at 60\% or 80\% of their full length, and participants predicted the final target. We tested the multimodal superiority and ocular primacy hypotheses, both of which were supported by the experiment.

In this paper we investigate the notion of legibility in sequential decision-making in the context of teams and teamwork. There have been works that extend the notion of legibility to sequential decision making, for deterministic and for stochastic scenarios. However, these works focus on one agent interacting with one human, foregoing the benefits of having legible decision making in teams of agents or in team configurations with humans. In this work we propose an extension of legible decision-making to multi-agent settings that improves the performance of agents working in collaboration. We showcase the performance of legible decision making in team scenarios using our proposed extension in multi-agent benchmark scenarios. We show that a team with a legible agent is able to outperform a team composed solely of agents with standard optimal behaviour.

As robot deployments become more commonplace, people are likely to take on the role of supervising robots (i.e., correcting their mistakes) rather than directly teaching them. Prior works on Learning from Corrections (LfC) have relied on three key assumptions to interpret human feedback: (1) people correct the robot only when there is significant task objective divergence; (2) people can accurately predict if a correction is necessary; and (3) people trade off precision and physical effort when giving corrections. In this work, we study how two key factors (robot competency and motion legibility) affect how people provide correction feedback and their implications on these existing assumptions. We conduct a user study ($N=60$) under an LfC setting where participants supervise and correct a robot performing pick-and-place tasks. We find that people are more sensitive to suboptimal behavior by a highly competent robot compared to an incompetent robot when the motions are legible ($p=0.0015$) and predictable ($p=0.0055$). In addition, people also tend to withhold necessary corrections ($p < 0.0001$) when supervising an incompetent robot and are more prone to offering unnecessary ones ($p = 0.0171$) when supervising a highly competent robot. We also find that physical effort positively correlates with correction precision, providing empirical evidence to support this common assumption. We also find that this correlation is significantly weaker for an incompetent robot with legible motions than an incompetent robot with predictable motions ($p = 0.0075$). Our findings offer insights for accounting for competency and legibility when designing robot interaction behaviors and learning task objectives from corrections.

A multiagent sequential decision problem has been seen in many critical applications including urban transportation, autonomous driving cars, military operations, etc. Its widely known solution, namely multiagent reinforcement learning, has evolved tremendously in recent years. Among them, the solution paradigm of modeling other agents attracts our interest, which is different from traditional value decomposition or communication mechanisms. It enables agents to understand and anticipate others' behaviors and facilitates their collaboration. Inspired by recent research on the legibility that allows agents to reveal their intentions through their behavior, we propose a multiagent active legibility framework to improve their performance. The legibility-oriented framework allows agents to conduct legible actions so as to help others optimise their behaviors. In addition, we design a series of problem domains that emulate a common scenario and best characterize the legibility in multiagent reinforcement learning. The experimental results demonstrate that the new framework is more efficient and costs less training time compared to several multiagent reinforcement learning algorithms.

One way to increase confidence in the outputs of Large Language Models (LLMs) is to support them with reasoning that is clear and easy to check -- a property we call legibility. We study legibility in the context of solving grade-school math problems and show that optimizing chain-of-thought solutions only for answer correctness can make them less legible. To mitigate the loss in legibility, we propose a training algorithm inspired by Prover-Verifier Game from Anil et al. (2021). Our algorithm iteratively trains small verifiers to predict solution correctness, "helpful" provers to produce correct solutions that the verifier accepts, and "sneaky" provers to produce incorrect solutions that fool the verifier. We find that the helpful prover's accuracy and the verifier's robustness to adversarial attacks increase over the course of training. Furthermore, we show that legibility training transfers to time-constrained humans tasked with verifying solution correctness. Over course of LLM training human accuracy increases when checking the helpful prover's solutions, and decreases when checking the sneaky prover's solutions. Hence, training for checkability by small verifiers is a plausible technique for increasing output legibility. Our results suggest legibility training against small verifiers as a practical avenue for increasing legibility of large LLMs to humans, and thus could help with alignment of superhuman models.