AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

Evaluating LLM-Based 0-to-1 Software Generation in End-to-End CLI Tool Scenarios

Ruida Hu, Xinchen Wang, Chao Peng, Cuiyun Gao, David Lo

The evolution of Large Language Models (LLMs) has catalyzed a paradigm shift towards intent-driven software development, where autonomous agents are expected to design and deliver complete, runnable software systems from scratch. However, existing benchmarks fail to adequately assess this 0-to-1 generation capability due to two fundamental limitations. First, they rely on predefined structural scaffolds, which reduces the task to mere file-filling. Second, they depend on rigid white-box unit testing, which forces generated code to conform to specific internal implementations rather than validating end-to-end user-centric behavior. To bridge this gap, we introduce CLI-Tool-Bench, a novel, structure-agnostic benchmark designed to evaluate the ground-up generation of Command-Line Interface (CLI) tools. Powered by an automated black-box differential testing framework, the benchmark comprises 94 high-quality, real-world repositories spanning diverse programming languages and complexity levels. For each task, agents are provided with an empty workspace, forcing them to autonomously handle repository planning and dependencies. We evaluate the generated software by executing it in isolated sandboxes. The system-level side effects and terminal outputs are then compared against human-written oracles using a rigorous multi-tiered equivalence metric. Extensive evaluation of seven state-of-the-art LLMs reveals that the top-tier models achieve a maximum overall success rate of only 43.8%, highlighting that 0-to-1 software generation remains a highly challenging frontier. Furthermore, we discover that agents exhibit a strong tendency to generate monolithic code structures, and that higher token consumption does not necessarily yield better task performance.

arXiv

What Models Express, Suppress, and Resist: Auditing Open-Weight LLMs with Persona Vectors

Winston Zeng, Ali Emami, Jinho D. Choi

What a language model will and will not do is largely set during post-training, but which behaviors it expresses, hides, or resists is not revealed by prompting alone. Persona vectors, behavioral directions in activation space, can probe this organization, but prior work covers only a handful of traits. We present the first systematic application of persona vectors at this scale, compiling a 53-trait inventory across four behaviorally distinct domains and labeling every trait in two open-weight models as natural (expressed at baseline), steerable latent but amplifiable, or intractable (resistant to standard extraction). Both models default to helpful, task-oriented behavior: all nine agentic traits are natural, and their default clinician behavior matches a board-certified psychologist's independent desirability judgments on 16 of 17 traits. Steering produces its largest gains on traits these defaults exclude: hyperbole, hallucination, and sycophancy. The same asymmetry holds across all 171 generic-trait pairs: two steerable traits can collapse the composition, but pairs involving a default never do. Where standard extraction fails on a trait like "evil," a vector transferred from a fine-tuned variant still recovers it, with the residual refusals appearing inside the model's chain-of-thought. Persona vectors are most informative not as a set of controls but as a probe of behavioral organization.

arXiv

A JoLT for the KV cache: Near-lossless KV cache compression via joint Lagrangian allocation of Tucker ranks and a rotated residual for llms

Rahul Krishnan, Volker Schulz

The key-value (KV) cache has become the dominant memory cost of transformer inference: it grows with batch size, context length, and depth, and at long context it, rather than the model weights, sets the throughput ceiling. Existing reductions fall into two families. Low-rank methods factor two-dimensional slices of the cache, either per-head matrices or cross-layer feature blocks, and quantization methods lower the bit-width of every entry. Neither exploits the fact that the cache at a layer is naturally a third-order tensor whose three axes, the heads, the tokens, and the features, carry very different amounts of redundancy. We take this tensor view directly. Our method, JoLT (Joint Lagrangian Tucker), applies a partial Tucker decomposition that compresses only the token and feature axes while leaving the head and layer axes intact, then restores the energy that truncation discards with a rotated low-bit residual: a random orthogonal rotation followed by low-bit quantization. A single Lagrangian dual allocates the Tucker ranks and the residual bit-widths together, per layer group and separately for keys and values, under one byte budget. The result is a near-lossless 2-3x compression. Perplexity stays near-lossless on both a grouped-query-attention model (Mistral-7B-v0.3) and a multi-head-attention model (LLaMA-2-13B), and GSM8K accuracy and needle-in-a-haystack retrieval hold at the uncompressed baseline at 2x on both architectures and through 3x on the GQA model. At 2x, JoLT reconstructs the cache to relative Frobenius error 0.009 (K) and 0.006 (V) on both architectures. A randomized-SVD variant, FlashJoLT, delivers a 5-13x compression-time speedup at 1024-token context and matched quality.

arXiv

Code-MUE: Measuring Code LLMs' Uncertainty through Execution-based Semantic Interaction Graphs

Xiaoning Ren, Yinxing Xue, Lei Ma, Yuheng Huang

As Code Large Language Models (LLMs) become central to modern software engineering, their inherent stochasticity poses significant real-world risks, where even minor errors can lead to severe functional, security, or safety consequences. Reliable automation, therefore, demands the ability to distinguish between confident, well-supported predictions and stochastic guessing. However, existing uncertainty estimation methods face a critical gap: white and grey-box techniques are often inapplicable to closed-source models, while standard "black-box" text metrics fail to capture the unique fragility of code, where syntactic variation does not always imply semantic divergence. To bridge this syntax-semantics gap, we introduce Code-MUE, a purely black-box framework that measures uncertainty through execution-based Semantic Interaction Graphs. Different from prior approaches that rely on superficial textual similarity, Code-MUE grounds uncertainty in observable runtime behavior, calculating the Von Neumann entropy of the solution space to quantify global semantic diversity. A large-scale empirical study across eight state-of-the-art LLMs demonstrates that Code-MUE achieves a strong negative correlation with functional correctness (Spearman's correlation up to -0.98), significantly outperforming lexical and embedding-based baselines while enabling robust risk detection and selective prediction in practical workflows.

arXiv

Scaling Point-in-Time Language Models

Bryan Kelly, Semyon Malamud, Johannes Schwab, Teng Andrea Xu

Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests and causal inference in finance and the social sciences. Point-in-time language models--trained exclusively on text available up to each calendar date--eliminate this leakage by construction, but existing efforts typically produce models that lag substantially behind their unconstrained counterparts. We show that this performance gap can be substantially narrowed through scale. Training decoder-only transformers with up to 4 billion parameters on 1 trillion chronologically filtered tokens from FineWeb, we construct a sequence of monthly model checkpoints spanning 2013-2024. Across a range of common-sense reasoning and language understanding benchmarks, our models approach the performance of leading open-weight models of comparable size (e.g., Gemma-3-4B and LLaMA-7B) trained on temporally unrestricted data, although a performance gap remains on several tasks. Instruction fine-tuning via LoRA further improves downstream usability. We release the complete pipeline--including dataset construction, training infrastructure, and evaluation code--to enable reproducible point-in-time language modeling and to support research applications that require strict temporal validity.

arXiv

PolyInterview: An LLM-based Platform for Immersive Mock Interview Practice with Comprehensive Multimodal Assessment

Zhiyuan Wen, Jiannong Cao, Kelly Chan, Zijian Wang, Chen Chen, Xiaoyun Liu, Jianing Yin, Zhuo Li

Preparing for job interviews is important for securing desired positions, yet realistic practice remains difficult to access: real interviews are infrequent, expert mock coaching is costly, and self-practice offers neither adaptive dialogue nor structured assessment. Existing systems typically address only parts of this need through fixed question sequences, limited communication channels, or feedback with little supporting evidence. We present PolyInterview, an LLM-based platform for immersive mock interview practice with comprehensive multimodal assessment. PolyInterview uses the target job description and CV to generate questions tailored to the role and candidate, conducts multi-turn spoken interviews with a lip-synced digital human interviewer that asks answer-aware follow-up questions, and evaluates response content, vocal delivery, and non-verbal behavior. Four parallel evaluators produce 13 behavior-level features that are aggregated into 10 assessment aspects and two competency tracks. Guided by the KSA and STAR frameworks, the report links each score to behavioral evidence and actionable recommendations. PolyInterview is publicly accessible. Its current all-account snapshot contains 101 accounts, 1,564 interview sessions, 7,665 generated questions, and 1,422 five-stage question sets. Generated questions are more closely aligned with their matched job description than with cross-role job descriptions in 93.7% of sessions. An evaluation by ten experts found strong question plans and actionable feedback.

arXiv

Conformal Graph Prediction with Z-Gromov-Wasserstein Distances

Gabriel Melo, Thibaut de Saivre, Anna Calissano, Florence d'Alché-Buc

Supervised graph prediction addresses regression problems where the outputs are structured graphs. Although several approaches exist for graph-valued prediction, principled uncertainty quantification remains limited. We propose a conformal prediction framework for graph-valued outputs, providing distribution-free coverage guarantees in structured output spaces. Our method defines nonconformity via the Z-Gromov-Wasserstein distance, instantiated in practice through Fused Gromov-Wasserstein (FGW), enabling permutation invariant comparison between predicted and candidate graphs. To obtain adaptive prediction sets, we introduce Score Conformalized Quantile Regression (SCQR), an extension of Conformalized Quantile Regression (CQR) to handle complex output spaces such as graph-valued outputs. We evaluate the proposed approach on a synthetic task and a real problem of molecule identification.

arXiv

Paired Uterine Whole-Slide Images and Pathology Reports for Multimodal Computational Pathology

Han Li, Jingsong Liu, Ayako Ura, Junlin Hou, Zhengyang Xu, Azar Kazemi, Oskar Thaeter, Christian Grashei, Fabian Gülhan, Reza Nasirigerdeh, Xun Ma, Rui Yan, Hao Chen, S. Kevin Zhou, Nassir Navab, Carolin Mogler, Peter Schüffler

Uterine diseases represent an important category of gynecologic pathology and require accurate histopathological assessment for diagnosis and treatment planning. Whole-slide images (WSI) have enabled the digital transformation of pathology workflows and provided new opportunities for artificial intelligence (AI) in computational pathology. In particular, multimodal models that jointly analyze histopathology images and pathology reports have shown promising potential for automated pathology report generation and AI-assisted diagnosis. However, the development of such systems remains limited by the scarcity of datasets that pair whole-slide images with clinically meaningful pathology reports. Instead, existing pathology datasets focus on patch- or slide-level annotations of a single endpoint (e.g., disease class), which do not fully capture the rich information in full clinical diagnostic workflow reports. Here, we introduce TUM-Uteria, a uterine pathology dataset comprising WSIs paired with diagnostic pathology reports at both the case and slide levels, collected from a tertiary medical center. The dataset contains 216 clinical cases, comprising 455 slide-level WSI-report pairs. The dataset underwent a structured multi-stage validation procedure involving board-certified pathologists to ensure reliable annotations. TUM-Uteria supports research in computational pathology, including whole-slide image analysis, multimodal learning, and automated pathology report generation.

arXiv

RePos: Relative-to-Absolute Pose Factorization for Cross-Environment WiFi-Based 3D Human Pose Estimation

Zhangcheng Hou, Tomoaki Ohtsuki

Device-free 3D human pose estimation from commodity WiFi Channel State Information (CSI) enables human sensing that preserves privacy and tolerates poor illumination, but its deployment is limited by poor generalization across environments. Unlike images, CSI measurements have no spatially localized correspondence to body parts and are heavily affected by multipath propagation. Consequently, models that regress absolute poses entangle body structure with location cues specific to each environment. Within a single environment this coupling is not problematic: RePos-D, a direct model that regresses the absolute pose, already achieves the best reported accuracy on Person-in-WiFi-3D, a 3.4% gain over the previous best WiFi method, DT-Pose. Across environments, however, the same model overfits position and degrades sharply. We therefore propose RePos, a factorized framework that separates root-relative pose estimation from root localization. By shielding the structure branch from absolute position, RePos learns robust pose representations. Specifically, it groups CSI features into latent tokens organized by body part that a skeleton-guided module refines into the pose, while a separate network estimates the root position from CSI amplitude through a differentiable spatial decomposition. Under the strict MM-Fi cross-environment protocol, RePos reduces the mean per-joint position error (MPJPE) by 10-21% over existing WiFi methods. The improvement is consistent across activity protocols, holds when each environment is held out in turn, and survives few-shot transfer without data leakage. Further analysis shows that the relative pose predictions remain largely independent of position, whereas root localization remains dependent on the environment.

arXiv

Orca: The World is in Your Mind

Yihao Wang, Yuheng Ji, Mingyu Cao, Yanqing Shen, Runze Xiao, Huaihai Lyu, Senwei Xie, Euan Liu, Klara Tian, Tianfeng Long, Yichi Zhang, Zhengliang Cai, Ruike Chen, Jifan Zhao, Ruochuan Shi, Zihan Tang, Jing Lyu, Wenxing Tan, Ningbo Zhang, Yangtao Hu, Yuming Gao, Xiansheng Chen, Junkai Zhao, Congsheng Xu, Boan Zhu, Ziqi Wang, Yupu Feng, Qiongqiong Zhang, Yingli Zhao, Yulong Ao, Shaoxuan Xie, You Liu, Guocai Yao, Leiduo Zhang, Xiaodan Liu, Yunyan Zhang, Yance Jiao, Xinyan Yang, Jiaxing Wei, Xu Liu, Tengfei Pan, Shaokai Nie, Chunlei Men, Sen Cui, Xiaojie Jin, Hongyang Li, Jianlan Luo, Yao Mu, Yunchao Wei, Jun Yan, Hang Zhao, Xiaolong Zheng, Jiaming Li, Yonghua Lin, Tiejun Huang, Zhongyuan Wang, Pengwei Wang

We introduce Orca, an initial instantiation of a general world foundation model. Orca learns a unified world latent space from multimodal world signals and exposes it through multimodal readout interfaces. Rather than optimizing isolated next-token, next-frame, or next-action prediction, we are centered on Next-State-Prediction modeling, offering a unified state-transition modeling route toward understanding, predicting, and acting upon the world. Orca learns through two complementary paradigms: unconscious learning captures dense natural state transitions from continuous videos, and conscious learning models sparse meaningful state transitions by language-described events and VQA supervision. For pre-training, we construct a large-scale world-learning inventory data, including 125K hours of video data and 160M event annotations. After pre-training, Orca learns a unified world latent space. To examine whether the learned latent supports downstream, we evaluate it by three representative downstream readouts: text generation, image prediction, and embodied action generation. Orca's backbone is frozen, and only the lightweight modality-specific decoders are trainable. Experiments show the scalability of the proposed paradigm and verify that stronger world latent enables stronger downstream readouts. Orca outperforms similar-sized specialized baselines. These results show that Orca, as a general world foundation model, presents a promising approach to understanding, predicting, and acting upon the world. Finally, we discuss the current limitations, aiming to provide useful insights and inspiration for the community.

arXiv

Reliability-Prioritized Fine-Grained Generation in Multimodal Large

Xiaomeng Fan, Wei Wu, Yuwei Wu, Zhi Gao, Shiyu Luo, Mingyang Gao, Haoyu Zhao, Zhenxin Diao, Yuxuan Ba, Lijia Feng, Yunde Jia, Mehrtash Harandi

Multimodal large language models (MLLMs) are increasingly expected to generate fine-grained descriptions of visual content. However, we observe and theoretically show that generating fine-grained responses poses a reliability challenge, \textit{i.e.}, fine-grained generation is more error-prone than coarse-grained generation. This phenomenon suggests that models should generate the finest description that remains reliable rather than simply produce more specific outputs. To investigate this problem, we develop \textsc{GranFact}, a granularity-aware benchmark consisting of expert-verified multi-object images with coarse-to-fine category annotations. Then, we design a hierarchy-aware evaluation algorithm, which assesses both whether model predictions are visually correct and how specific the correct predictions are. We also propose a reliability-prioritized preference optimization method based on Direct Preference Optimization, which penalizes unreliable fine-grained claims while rewarding reliable specificity. Experiments on \textsc{GranFact} show that our method improves fine-grained generation while preserving reliability. Code and data are available \href{https://github.com/WeiWu2025/GranFact}{here}.

arXiv

Depth-Semantic Alignment and Affinity-Guided Fusion for Structured Radar Point Cloud Generation

Amjad Hussain, Wenjie Liu, Yuchen Tan, Fuyuan Ai, Zecheng Li, Chunyi Song, Xin Qiu

Point clouds are an important carrier of three-dimensional spatial information, and their quality directly affects the performance of downstream perception tasks such as object detection and tracking. However, millimeter-wave radar point clouds are typically sparse, noisy, and structurally incomplete. To address these limitations, this paper proposes a multimodal point cloud generation method based on vision-radar fusion. The proposed method leverages image semantic information to impose structural constraints and achieve spatial alignment for radar point clouds, while incorporating a sparse completion strategy to enhance point density and recover missing structures. The generated point clouds are further evaluated in object detection and tracking tasks. Experimental results demonstrate that the proposed method effectively improves point cloud quality and enhances the detection accuracy and robustness of perception models in complex environments, providing a practical solution for multisensor point cloud generation and intelligent perception systems.

arXiv

Where Will They Go? Modelling Multimodal Pedestrian Manoeuvres from Ego-centric Videos

Yuxuan Xie, Nicolas Pugeault, Chongfeng Wei, Hubert P. H. Shum, Edmond S. L. Ho

Pedestrian trajectory prediction from an on-board ego-centric camera is challenging since it depends on complex interactions with vehicles and scene context, as well as the intention of the pedestrian. The task becomes even more challenging since pedestrian intention is often ambiguous from historical observations alone, leading to an inherently multimodal distribution over future trajectories. Existing CVAE-based predictors learn a latent representation by aligning the prior conditioned on trajectory history with the posterior conditioned on both history and future trajectories. This alignment may suppress multimodality in the latent space, leading to mixed-mode predictions that interpolate between distinct future behaviours. In this paper, we propose MMPM, a mode-aware framework that separately models future trajectory distributions into semantically meaningful modes based on the pedestrian's crossing behavior. MMPM consists of two modules: behavior-aware Pedestrian Interaction Module (PIM) that jointly captures pedestrian-vehicle and pedestrian-environment interactions by introducing gaze and hand gestures, and a CVAE-based Mode-aware Trajectory Predictor (MTP) module to model the future trajectory distributions in two modes, crossing and non-crossing the road, separately. A query-based decoder further enforces mode consistency during decoding. Experiments on PIE and JAAD datasets show that our method surpasses state-of-the-art baselines. Our proposed MTP is model-agnostic, which can be integrated into existing frameworks such as BiTrap and SGNet to further improve future trajectory prediction performance. We additionally introduce a data-driven validation protocol that retrieves spatio-temporally consistent real-world trajectories from the same scene and evaluates predictions against these retrieved examples, demonstrating improved frame-wise displacement errors over previous work.

arXiv

Interaction-Aware Whole-Body Control for Compliant Object Transport

Hao Zhang, Yves Tseng, Ding Zhao, H. Eric Tseng

Cooperative object transport in unstructured environments remains challenging for assistive humanoids because strong, time-varying interaction forces can make tracking-centric whole-body control unreliable, especially in close-contact support tasks. This paper proposes a bio-inspired, interaction-oriented whole-body control (IO-WBC) that functions as an artificial cerebellum - an adaptive motor agent that translates upstream (skill-level) commands into stable, physically consistent whole-body behavior under contact. This work structurally separates upper-body interaction execution from lower-body support control, enabling the robot to maintain balance while shaping force exchange in a tightly coupled robot-object system. A trajectory-optimized reference generator (RG) provides a kinematic prior, while a reinforcement learning (RL) policy governs body responses under heavy-load interactions and disturbances. The policy is trained in simulation with randomized payload mass/inertia and external perturbations, and deployed via asymmetric teacher-student distillation so that the student relies only on proprioceptive histories at runtime. Extensive experiments demonstrate that IO-WBC maintains stable whole-body behavior and physical interaction even when precise velocity tracking becomes infeasible, enabling compliant object transport across a wide range of scenarios.

arXiv

Inelastic Constitutive Kolmogorov-Arnold Networks: A generalized framework for automated discovery of interpretable inelastic material models

Chenyi Ji, Kian P. Abdolazizi, Hagen Holthusen, Christian J. Cyron, Kevin Linka

A key problem of solid mechanics is the identification of the constitutive law of a material, that is, the relation between strain history and stress. Machine learning has lead to considerable advances in this field lately. Here we introduce inelastic Constitutive Kolmogorov-Arnold Networks (iCKANs). This novel artificial neural network architecture can discover in an automated manner symbolic constitutive laws describing both the elastic and inelastic behavior of materials. That is, it can translate data from material testing into corresponding elastic and inelastic potential functions in closed mathematical form. We demonstrate the advantages of iCKANs using both synthetic data and experimental data of the viscoelastic polymer materials VHB 4910 and VHB 4905. The results demonstrate that iCKANs accurately capture complex viscoelastic behavior while preserving physical interpretability. It is a particular strength of iCKANs that they can process not only mechanical data but also arbitrary additional information available about a material (e.g., about temperature-dependent behavior). This makes iCKANs a powerful tool to discover in the future also how specific processing or service conditions affect the properties of materials.

arXiv

UCOB: Learning to Utilize and Evolve Agentic Skills via Credit-Aware On-Policy Bidirectional Self-Distillation

Songjun Tu, Chengdong Xu, Qichao Zhang, Yiwen Ma, Yaocheng Zhang, Linjing Li, Dong Li, Xiangyuan Lan, Dongbin Zhao

Skill memories can improve agentic reinforcement learning by reusing past experience as textual guidance, but retrieved skills are not oracular: they may help in one state while misleading the same policy in another. This makes the common privileged-teacher assumption fragile, namely that a skill-conditioned prompt can be treated as a fixed teacher for the no-skill prompt. We introduce UCOB, a framework for learning to utilize and evolve agentic skills via credit-aware on-policy bidirectional self-distillation. UCOB treats skill-conditioned and no-skill prompts as two on-policy context views of the same model, compares their return-to-go within the same task and anchor state, and uses the higher-return view as the local teacher. This local credit signal internalizes useful skill-conditioned behavior, corrects misleading skill usage, and guides task/state skill memory updates, utility-aware retrieval, and reflection self-training. Experiments on agentic tasks, including ALFWorld, WebShop, and Search-QA, show that UCOB outperforms skill-free RL, skill-memory baselines, and self-distillation methods across model scales, with up to 23.5 and 18.0 point gains over SOTA baselines on ALFWorld and WebShop. Ablations and analyses further validate its core mechanisms, continual adaptation across environments, and modest training overhead. Code is available at https://github.com/TU2021/UCOB.

arXiv

An expressivity analysis of hierarchical modelling in deep transformers via bounded-depth grammars

Vinoth Nandakumar, Qiang Qu, Pramod Thebe, Sakshi Khachariya, Tongliang Liu

Deep neural networks are widely believed to derive their expressive power from their ability to form \textbf{hierarchical representations}, capturing progressively more abstract and compositional features across layers. In language modeling, \textbf{transformers} have emerged as the dominant architecture, with early layers capturing local syntactic patterns and later layers encoding more complex clause-level dependencies. While this intuition has shaped model design, there remains a lack of rigorous theoretical work demonstrating \textbf{how} deep transformers represent such hierarchical structures. In this work, we analyze the expressiveness of deep transformer models through the formal lens of bounded-depth, non-recursive context-free grammars. For this class of grammars, we explicitly construct transformers with positional attention whose depth grows linearly with grammar depth, while the neuron count scales with the number of derivation-tree shapes and quadratically with the number of production rules. Our theoretical results support the linear representation hypothesis by demonstrating that these architectures possess the structural capacity to encode abstract grammatical states into low-dimensional, linearly separable subspaces within the residual stream.

arXiv

AuAu: A Benchmark for Auditing Authoritarian Alignment in Large Language Models

Andreas Einwiller, Max Klabunde, Florian Lemmerich

The worldwide rise of authoritarianism and the growing role of Large Language Models (LLMs) in users' everyday lives raise the question of whether specific models exhibit or promote authoritarian attitudes. We introduce AuAu, a comprehensive benchmark for assessing the risk of authoritarian tendencies in LLM responses. AuAu combines three evaluation approaches: (i) psychometric questions from 15 human-validated instruments, (ii) vignettes probing intended behavior in concrete situations, and (iii) responses to realistic user prompts. Unlike prior work, AuAu measures not only overall authoritarian alignment but also its established sub-concepts: Authoritarian Aggression, Authoritarian Submission, and Conventionalism. Evaluating 17 models from China, the EU, Russia, and the USA, we find substantial authoritarian response rates on psychometric instruments across all models, though rates drop significantly on more realistic downstream tasks. Moreover, a simple authoritarian system prompt manipulates 15 of 17 models into promoting increased authoritarianism. Our results underscore the need for continued, systematic auditing of LLM-based AI systems to detect and mitigate authoritarian tendencies in their outputs.

arXiv

It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO

Naihao Deng, Yilun Zhu, Naichen Shi, Clayton Scott, Rada Mihalcea

Warning: This paper contains several toxic and offensive statements. Modern large language models (LLMs) are typically aligned through large-scale post-training to ensure fair and reliable behavior. In this work, we investigate how easily such guardrails can be broken by Group Relative Policy Optimization (GRPO). We show that one-shot GRPO training on a single biased example is sufficient to induce systematic bias, with stereotype-driven reasoning generalizing across attributes, categories, and benchmarks. We further find that models differ in their susceptibility based on the initial likelihood of producing biased outputs. Our results reveal a critical vulnerability in post-training: alignment can be overridden by a single example.

arXiv

What Should a Skill Remember? Quality--Cost Trade-offs in Cost-Aware Skill Rewriting for Language Model Agents

Qinghua Xing, Yinda Chen, Yaping Jin, Zhenhe Wu, Bohan Lin, Hang Zhou, Xinghao Chen, Hanting Chen, Zhiwei Xiong

Large language model agents increasingly rely on skills: reusable procedural documents encoding workflows, tool use, implementation patterns, validation checks, and domain rules. Skill rewriting is often treated as prompt compression, but shorter skills can make agents more expensive by removing sparse operational anchors that prevent exploration, debugging, and recovery. We study skill rewriting through this economic lens. Our controlled framework profiles skill structure, rewrites skills using information-preservation strategies, and evaluates the rewrites under fixed task instructions, environments, and verifiers. Experiments on SkillsBench reveal distinct quality--cost trade-offs across strategies: API/code anchoring, workflow guarding, and rule/formula anchoring benefit different task families, with no universally dominant template. In the main held-out evaluation, the learned policy reduces total cost by 7.0% and downstream agent-token cost by 6.0%; in frozen cross-model transfer, the corresponding reductions average 14.7% and 13.7%, while verifier quality is preserved. These results position skill design as cost-aware operational knowledge engineering rather than prompt compression. Resources: https://github.com/1Reminding/Skill_EE.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.