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

Relay-Bench: Evaluating LLMs on Multi-Domain Reasoning Chains

Liam Swayne

Introducing Relay-Bench, an unsaturated, holistic, text-only benchmark that measures LLMs' ability to complete an assortment of tasks from distinct domains in a single prompt. The leading model, GPT-5.5 (xHigh), scores 43.3%. The test set entirely consists of composite problems: groups of single-domain subproblems that are strung together into challenges that require reasoning across multiple domains in combination. Many of these problems then have layers of complexity added through prompt encoding and deliberate context bloat. Domains tested include visual reasoning, coding, math, information extraction (with a focus on web search), problem-solving, general knowledge, and data analysis. No restrictions are imposed outside of the model harness, and models are explicitly encouraged to leverage code-execution, web searches, and all available tools. All problems are composed of two to thirteen subproblems and do not require multi-modal input or output.

arXiv

Building a European Multilingual Evaluation Dataset: The MMLU Localisation Project within the EMT Network

Pilar Sánchez-Gijón, Susana Valdez, Sofía Calvo Del Barrio, Florence Bellemont, Anna Kokkinidou, Mihai Cristian Brasoveanu

This paper reports on a collaboration between the Directorate-General for Translation (DGT) and the European Master's in Translation (EMT) to localise the MMLU dataset into 11 European languages. Beyond creating a more inclusive benchmark for LLM evaluation, the project offers master's students authentic, project-based professional training in translation, revision, project management, and multilingual coordination, while highlighting key methodological, administrative, and workflow challenges.

arXiv

Enabling Multilingual Privacy Policy Audits: Large-Scale Analysis of Spanish Mobile Apps

Marcos Moran, David Rodriguez, Luka Nenadic, Norman Sadeh, Jose M. Del Alamo

Automated analyses of privacy policies enable large-scale assessments of transparency in digital ecosystems, yet existing auditing pipelines remain predominantly English-centric. This limits their ability to systematically evaluate multilingual environments, as in the European Union, where many services disclose privacy practices only in local languages. This paper examines whether large language models (LLMs) can extend privacy policy analysis beyond English without requiring language-specific adaptation, thus empowering large-scale auditing in linguistically diverse app ecosystems. We assemble an evaluation corpus spanning all 24 official EU languages from translated versions of two established expert-annotated datasets (OPP-115 and MAPP) and assess translation fidelity through automated metrics and targeted legal-expert review. Our LLM-based classifier for identifying categories of personal data collection achieves stable cross-lingual performance, with macro-F1 scores ranging between 0.91 and 0.94. We then leverage this capability in a large-scale audit of 2,611 Android applications from the Spanish Google Play Store. Combining multilingual privacy policy analysis with the evaluation of corresponding privacy labels and runtime network traffic exposes an important linguistic barrier: public-sector apps predominantly provide privacy policies in Spanish, whereas popular commercial apps mostly provide them in English. We reveal systematic discrepancies between declared and observed practices, especially in public-sector apps. Overall, our results indicate how English-only privacy audits can systematically obfuscate transparency gaps in multilingual environments.

arXiv

Convolution for Large Language Models

Yuchuan Tian, Yingte Shu, Wei He, Shuo Zhang, Tianchen Zhao, Chao Xu, Xinghao Chen, Yunhe Wang, Hanting Chen, Yu Wang

Large language models (LLMs) largely rely on Transformers, where self-attention provides global token interaction but does not explicitly encode the locality of natural language. We study whether lightweight depthwise convolutions can supply this local inductive bias without materially increasing model size. Our macro-level ablation compares convolution at 17 locations in a Qwen3 Transformer block and finds the best results when convolution is applied to the projected queries, keys, and values before attention. A subsequent micro-level study favors a residual depthwise convolution with kernel size $k=3$, without additional normalization or activation. Across Qwen3 models and several pre-training data budgets, this design improves the average accuracy on seven downstream benchmarks while adding less than $0.01\%$ parameters. A representation-level case study further suggests that the convolution makes repeated token IDs more sensitive to their immediate context. These results support depthwise convolution as a lightweight complement to self-attention for modeling short-range token interactions.

arXiv

Operational Hallucination and Safety Drift in AI Agents

Shasha Yu, Fiona Carroll, Barry L. Bentley

Large language models (LLMs) serving as planners in tool-using autonomous agents introduce dynamic reliability risks in multi-turn execution. While single-turn safety mechanisms are relatively mature, extended interactions reveal structural vulnerabilities where initial alignment degrades over time. This paper empirically characterizes two observed failure modes across multiple state-of-the-art LLMs: Safety Drift, the gradual erosion of declared safety intent leading to constraint-violating actions (e.g., textual refusal followed by reconnaissance and unsafe execution), and Operational Hallucination, persistent repetitive tool calls indicative of flawed state perception (e.g., livelocks even in legitimate tasks). Through controlled multi-turn evaluation on high-stakes ethical dilemmas, malicious requests, and benign controls, we quantify these phenomena using declaration-action gap and livelock metrics, demonstrating their cross-model prevalence under direct execution protocols. Root-cause analysis attributes the instabilities to the decoupling of reasoning context from execution state in current agent loops. We propose an Action-Aware Supervision Layer - a lightweight, plug-and-play architectural blueprint incorporating intent-action consistency checks, runtime state tracking, and forced termination primitives. Post-hoc simulation on captured failure trajectories shows the layer can intercept observed violations without false positives on benign cases. This work advances agent reliability by shifting focus from linguistic safeguards to enforceable architectural mechanisms for responsible agentic AI.

arXiv

A Controlled Study of Attention-Only Transformers

Henry Ndubuaku, Karen Mosoyan, Jakub Mroz, Noah Cylich, Satyajit Kumar, Parkirat Sandhu, Roman Shemet, Justin H Lee

Feed-forward networks hold two thirds of a transformer's non-embedding parameters, yet the architecture has not received a necessity test that controls parameters, compute, and depth at once. We pretrain attention-only decoder transformers (Simple Attention Networks, SANs) against standard transformers matched separately for parameter count, training FLOPs, and depth (2 to 48 layers), for up to 105B tokens at 6M to 87M parameters. Deleting feed-forward layers in place is costly: the standard transformer leads by 0.47 nats at matched depth and 0.26 nats at matched FLOPs. Reallocating the freed budget into attention depth closes the gap: at matched parameters the difference is 0.006 nats (0.27 percent of loss), reproducible to one part in ten thousand across seed pairs, shrinking across 5B, 30B, and 105B budgets, and holding near 0.02 nats across a 29x size range. Three measurements localize the remaining gap to parametric recall: attention-only models are better on context-grounded answers and worse where knowledge must come from weights. Weight spectra show why: routing matrices (Q/K) crystallize early, content matrices accumulate rank slowly, and removing feed-forward layers relocates this accumulation to the attention output projection. QK-normalization, not feed-forward layers or residual gating, keeps 48-layer attention-only stacks trainable. The deficit concentrates on low-context query prediction and localizes there entirely by the largest budget. A pre-registered test confirms the account: it predicts a 0.02 to 0.05 nat gap on knowledge-dense web text; a matched pair trained on fineweb-edu measures 0.040. Within the tested regime, attention does the rest.

arXiv

SpecXMaster Technical Report

Yutang Ge, Yaning Cui, Hanzheng Li, Jun-Jie Wang, Fanjie Xu, Jinhan Dong, Yongqi Jin, Dongxu Cui, Peng Jin, Guojiang Zhao, Hengxing Cai, Tianci Yangfeng, Xueqing Chen, Hongshuai Wang, Rong Zhu, Linfeng Zhang, Xiaohong Ji, Zhifeng Gao

Intelligent spectroscopy serves as a pivotal element in AI-driven closed-loop scientific discovery, functioning as the critical bridge between matter structure and artificial intelligence. However, conventional expert-dependent spectral interpretation encounters substantial hurdles, including susceptibility to human bias and error, dependence on limited specialized expertise, and variability across interpreters. To address these challenges, we propose SpecXMaster, an intelligent framework leveraging Agentic Reinforcement Learning (RL) for NMR molecular spectral interpretation. SpecXMaster enables automated extraction of multiplicity information from both 1H and 13C spectra directly from raw FID (free induction decay) data. This end-to-end pipeline enables fully automated interpretation of NMR spectra into chemical structures. It demonstrates superior performance across multiple public NMR interpretation benchmarks and has been refined through iterative evaluations by professional chemical spectroscopists. We believe that SpecXMaster, as a novel methodological paradigm for spectral interpretation, will have a profound impact on the organic chemistry community.

arXiv

Precision-Varying Prediction (PVP): Robustifying ASR systems against adversarial attacks

Matías Pizarro, Raghavan Narasimhan, Jonas Killian, Asja Fischer

With the increasing deployment of automated and agentic systems, ensuring the adversarial robustness of automatic speech recognition (ASR) models has become highly relevant. We observe that changing the precision of an ASR model during inference reduces the likelihood of adversarial attacks to succeed. We take advantage of this fact to make models more robust simply by randomly sampling the precision during prediction. Moreover, this insight can be turned into an adversarial example detection strategy by implementing a simple Gaussian classifier that thresholds the differences between outputs of models run with different precision. To further enhance security boundaries, we combine the approach with an existing uncertainty-based defense mechanism, which forces adaptive adversaries to introduce highly perceptible noise to bypass detection. An experimental analysis across various ASR models, languages, and attack types demonstrates a significant increase in adversarial robustness, competitive detection capabilities, and resistance to adaptive threats.

arXiv

SkillRouter: Skill Routing for LLM Agents at Scale

YanZhao Zheng, ZhenTao Zhang, Chao Ma, YuanQiang Yu, JiHuai Zhu, Yong Wu, Tianze Xu, Baohua Dong, Hangcheng Zhu, Ruohui Huang, Gang Yu

Reusable skills let LLM agents package task-specific procedures, tool affordances, and execution guidance into modular building blocks. As skill ecosystems grow to tens of thousands of entries, exposing every skill at inference time becomes infeasible. This creates a skill-routing problem: given a user task, the system must identify relevant skills before downstream planning or execution. Existing agent stacks often rely on progressive disclosure, exposing only skill names and descriptions while hiding the full implementation body. We examine this design choice on a SkillsBench-derived benchmark with approximately 80K candidate skills, targeting the practically important setting of large skill registries with heavy overlap. Across representative dense and reranking baselines on this setting, hiding the skill body causes a 37-44 percentage point drop in routing accuracy. Stronger controls show that the missing signal is body-resident rather than a simple length artifact: body-distilled descriptions recover part of the gap, but remain 7-21 points below direct all-field routing, while a metadata-only encoder trained with the same data remains 14.0 points below its all-field counterpart. Motivated by this finding, we present Skillrouter, a compact 1.2B body-aware retrieve-and-rerank pipeline. Skillrouter achieves 74.0% Hit@1 on our benchmark -- the strongest average top-1 routing performance among the baselines we evaluate -- while using 13$\times$ fewer parameters and running 5.8$\times$ faster than the strongest base pipeline. The ranking gains further generalize to a supplementary benchmark independently constructed from three skill sources. In a complementary end-to-end study across four coding agents, routing gains transfer to improved task success, with larger gains for more capable agents.

arXiv

Beyond Access: Guided LLM Scaffolding for Independent Learning in Undergraduate Statistics

Mohammad Amanlou, Yasaman Amou-Jafari, Mehrad Livian, Fatemeh Boloukazari, Fereshte Bagheri, Elahe Khodaverdi Nadrabadi, Shahab Sherafat, Behnam Bahrak

Large language models (LLMs) are increasingly entering students' learning practices, but their educational value depends on whether they support reasoning or enable task completion without engagement. This study examines guided LLM use in an undergraduate Probability and Statistics course, focusing on the gap between assigned access and actual interaction quality. In a four-week quasi-experimental summer program, students were organized into three balanced conditions: no LLM access, unrestricted LLM access, and guided LLM access. The guided condition used the same LLM platform as the unrestricted condition, but students received explicit training and rules promoting reasoning-focused help-seeking, stepwise hints, verification, and ethical use. All quizzes and the delayed final exam were completed without LLM or external assistance, allowing us to distinguish AI-supported practice performance from independent learning. Results show that guided use was associated with clearer learning-oriented interaction patterns than unrestricted access, especially in prioritizing reasoning over final answers and requesting stepwise support. Guided-LLM students showed stronger no-help quiz performance during the intervention phase, whereas unrestricted access appeared more useful for assisted practice completion than for consistently improving independent performance. Available time measures did not support a simple duration-based explanation, and self-assessment calibration suggested better alignment between perceived and demonstrated understanding in the Guided-LLM condition. Overall, LLM access alone appears to be an incomplete educational intervention. For Artificial Intelligence in Education (AIED), the central design challenge is to scaffold how students use LLMs so that these systems function as partners in reasoning rather than answer-getting tools.

arXiv

A Classifier That Teaches Itself: Self-Improving, Frozen-gate Training (SIFT) for Dynamic Document Classification

Bogdan Raduta, Horia Velicu, Alexandru Preda, Serban Chiricescu

Document classification is a solved problem in the laboratory and an unsolved one in the enterprise. The blocker is rarely model architecture; it is the labeling project that must precede a model and the institutional fear of letting a model retrain itself once one exists. We present SIFT (Self-Improving, Frozen-gate Training), a dynamic classifier service, which attacks both. SIFT serves classification from a deliberately cheap, CPU-bound pipeline, a SPLADE sparse encoder feeding a LightGBM head, and escalates only the low-confidence minority of pages to an LLM judge. The judge's verdicts are written back into a labeled corpus, so the expensive model continuously teaches the cheap one: the escalation rate falls, the corpus grows from production traffic rather than from an up-front annotation effort, and accuracy compounds with use. Onboarding a new document family requires only a declarative bundle, label space, anchor phrases, and a judge glossary, not a labeling project. The harder problem is safety: an autonomously retraining classifier can silently regress. SIFT resolves this with a two-part promote gate, a critical-label F1 regression check plus a frozen golden regression set the model is never trained on, either of which vetoes promotion. This turns "retrain monthly without a human" from reckless into routine. We describe the architecture, the self-feeding corpus loop, the frozen-gate promotion mechanism, and an illustrative multi-domain deployment, and we discuss the economics of a classifier whose marginal labeling cost trends toward zero.

arXiv

Stochastic Resetting Accelerates Reinforcement Learning Beyond Random Search

Jello Zhou, David J. Schwab, Vudtiwat Ngampruetikorn

Stochastic resetting -- intermittently returning a process to a fixed reference state -- has emerged as an effective mechanism for optimizing first-passage properties. Existing theory largely treats processes that search but do not learn: the searcher follows fixed dynamics, accumulating no knowledge between resets. Here we ask how stochastic resetting interacts with reinforcement learning, where the underlying dynamics adapt through experience. In tabular grid environments, we find that resetting can accelerate learning even when it does not reduce the search time of a diffusive agent. Our results reveal a distinct additional mechanism through which resetting speeds the propagation of reward information. We show that deterministic, sharp resetting accelerates learning more than the stochastic protocol but over a narrower range of reset rates. In a continuous-state task with neural-network-based value approximation, we demonstrate that resetting speeds up learning when exploration is hard and rewards are sparse. We argue further that, in the tabular tasks, resetting accelerates learning without altering the solution the agent ultimately reaches, unlike other techniques such as temporal discounting, which biases the optimal behavior. Our results establish stochastic resetting as a simple, tunable mechanism for accelerating learning by shaping how experience accumulates, extending a canonical phenomenon of statistical mechanics to adaptive systems.

arXiv

Benchmarking Multimodal Large Language Models for Scientific Visualization Literacy

Patrick Phuoc Do, Chau M. Ta, Chaoli Wang

Multimodal large language models (MLLMs) are increasingly used to interpret visualizations, yet current evaluations remain largely chart-centric and provide limited evidence of understanding of scientific visualization (SciVis). We benchmark six MLLMs on the scientific visualization literacy assessment test, a standardized SciVis literacy assessment comprising 49 items based on 18 scientific visualizations and illustrations, spanning 8 techniques and 11 task types. We evaluate three closed-source and three open-source models under a closed-world protocol and compare their performance using data from 485 human participants. Results show that current MLLMs do not exhibit uniform SciVis literacy. Gemini is the strongest model overall, exceeding the human mean across the evaluated subsets, whereas the open-source models remain below the human baseline. Performance is highly uneven across techniques and tasks: models perform best on scientific illustration, search, and spatial understanding, but struggle on texture-based and integration-based visualizations and on quantitative estimation. Error analysis reveals recurring failures in fine-grained quantitative estimation, flow-direction interpretation, and grounded encoding interpretation. These findings position SciVis literacy as a necessary benchmark dimension for evaluating multimodal AI systems. Our code and model outputs are publicly available at https://github.com/patdmp/mllm-scivis-lit-benchmark.

arXiv

AI for Auto-Research: Roadmap & User Guide

Lingdong Kong, Xian Sun, Wei Chow, Linfeng Li, Kevin Qinghong Lin, Xuan Billy Zhang, Song Wang, Rong Li, Qing Wu, Wei Gao, Yingshuo Wang, Shaoyuan Xie, Jiachen Liu, Leigang Qu, Shijie Li, Lai Xing Ng, Benoit R. Cottereau, Ziwei Liu, Tat-Seng Chua, Wei Tsang Ooi

AI-assisted research is crossing a threshold: fully automated systems can now generate research papers for as little as $15, while long-horizon agents can execute experiments, draft manuscripts, and simulate critique with minimal human input. Yet this productivity frontier exposes a deeper integrity problem: under scientific pressure, even frontier LLMs still fabricate results, miss hidden errors, and fail to judge novelty reliably. Studying developments through April 2026, we present an end-to-end analysis of AI across the complete research lifecycle, organized into four epistemological phases: Creation (idea generation, literature review, coding & experiments, tables & figures), Writing (paper writing), Validation (peer review, rebuttal & revision), and Dissemination (posters, slides, videos, social media, project pages, and interactive agents). We identify a sharp, stage-dependent boundary between reliable assistance and unreliable autonomy: AI excels at structured, retrieval-grounded, and tool-mediated tasks, but remains fragile for genuinely novel ideas, research-level experiments, and scientific judgment. Generated ideas often degrade after implementation, research code lags far behind pattern-matching benchmarks, and end-to-end autonomous systems have not yet consistently reached major-venue acceptance standards. We further show that greater automation can obscure rather than eliminate failure modes, making human-governed collaboration the most credible deployment paradigm. Finally, we provide a structured taxonomy, benchmark suite, and tool inventory, cross-stage design principles, and a practitioner-oriented playbook, with resources maintained at our project page.

arXiv

Adaptive Multi-Round Allocation with Stochastic Arrivals

Yuqi Pan, Davin Choo, Haichuan Wang, Milind Tambe, Alastair van Heerden, Cheryl Johnson

We study a sequential resource allocation problem motivated by adaptive network recruitment, in which a limited budget of identical resources must be allocated over multiple rounds to individuals with stochastic referral capacity. Successful referrals endogenously generate future decision opportunities while allocating additional resources to an individual exhibits diminishing returns. We first show that the single-round allocation problem admits an exact greedy solution based on marginal survival probabilities. In the multi-round setting, the resulting Bellman recursion is intractable due to the stochastic, high-dimensional evolution of the frontier. To address this, we introduce a population-level surrogate value function that depends only on the remaining budget and frontier size. This surrogate enables an exact dynamic program via truncated probability generating functions, yielding a planning algorithm with polynomial complexity in the total budget. We further analyze robustness under model misspecification, proving a multi-round error bound that decomposes into a tight single-round frontier error and a population-level transition error. Finally, we evaluate our method on real-world inspired recruitment scenarios.

arXiv

Belief-reality separation lives in routing over a shared value slot in language models

Oliver Steele, Jiangtao Wen, Yuxing Han

Capable language models hold what a character believes apart from what is true: told "Anna believes the cup is blue; in reality it is red," they answer blue about Anna and red about the world. Where in the computation does that separation live? We show it rests on two separable mechanisms at two positions. A generic value slot binds the attributed value. A router at the query position selects which frame, the character's belief or reality, a query reads out. Two routes fill the slot: an asserted belief, whose value the text supplies, binds in directly; a derived belief, whose value must be inferred from what the character could see, arrives by a visibility-gated lookback. A subspace trained on either route steers the other, and only the derived route depends on described visibility. The slot itself carries no belief-reality tag: intervening on it moves a reality readout as strongly as a belief one. The separation lives instead in a dissociated pair of routing subspaces, which flip a query between frames without injecting the donor's value. These results hold across three architectures, on stimuli de-confounded against theory-of-mind-benchmark shortcuts; the behavior itself emerges between 3B and 7B across five model families. This paper develops the single belief-reality axis in depth; a companion paper shows the same slot-and-router format is shared across the other non-actual contexts a sentence can open (counterfactual, fictional, temporal).

arXiv

Subliminal Clocks: Latent Time Modelling in Diffusion Language Models

Maximo Eduardo Rulli, Thomas Vaitses Fontanari, Simone Petruzzi, Federico Alvetreti, Giorgio Strano, Donato Crisostomi, Giorgos Nikolaou, Tommaso Mencattini, Andrea Santilli, Emanuele Rodolà, Simone Scardapane, Alessio Devoto

Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive models. Unlike standard diffusion-based approaches, DLMs are not explicitly conditioned on a timestep, raising a natural question: do these models internally represent denoising progress, and how is such information used downstream? In this work, we show that DLMs do in fact encode a latent representation related to the diffusion timestep within their residual streams. We find that this signal can be reliably extracted using probes across layers, indicating that denoising progress is decodable from internal activations. We further demonstrate that steering the model along a low-dimensional subspace associated with the inferred timestep allows us to systematically modulate its notion of denoising progress, leading to predictable changes in model confidence and entropy. Finally, we analyse the geometry of the identified representation, showing that it exhibits structured and interpretable properties in activation space, and shedding light on how such a signal is processed by these models.

arXiv

Routing by Reaching: Composition of Pre-trained GFlowNets for Multi-Objective Generation

Seokwon Yoon, Youngbin Choi, Seunghyuk Cho, Seungbeom Lee, MoonJeong Park, Dongwoo Kim

Generative Flow Networks (GFlowNets) learn to sample diverse candidates in proportion to a reward function, making them well-suited for scientific discovery, where exploring multiple promising solutions is crucial. Further extending GFlowNets to multi-objective settings has attracted growing interest as real-world applications often involve multiple, conflicting objectives. However, existing approaches require joint training for each combination of objectives, meaning that any change in the objective set necessitates retraining from scratch. We propose a framework that composes pre-trained GFlowNets at inference time, enabling rapid adaptation without fine-tuning or retraining. Importantly, our framework is flexible, capable of handling diverse reward combinations ranging from linear scalarization to complex nonlinear operators, which are often handled separately in previous literature. We prove that our method exactly recovers the target distribution for linear scalarization, and quantify the approximation quality for nonlinear operators through a distortion factor. Experiments on a synthetic 2D grid and real-world molecule generation tasks demonstrate that our approach achieves performance comparable to baselines.

arXiv

Rethinking Efficiency in Neural Combinatorial Optimization: Batched Preference Optimization with Mamba

Zhenxing Xu, Zeyuan Ma, Weidong Bao, Yan Zheng, Chongshuang Hu, Ji Wang, Zhiguang Cao

We study efficiency as a first-class objective in Neural Combinatorial Optimization (NCO) and present ECO, an efficient learning framework that combines batched preference optimization with a Mamba backbone. Instead of tightly interleaving every policy update with on-policy rollouts, ECO decouples trajectory generation from gradient updates through two stages: supervised warm-up on pre-computed solutions and iterative Direct Preference Optimization (DPO) on batched candidate sets generated by the current policy. We pair this learning pipeline with a mixed Mamba encoder-decoder that reduces memory growth on long sequences and improves hardware utilization. A local-search-guided bootstrapping strategy is further used during training to widen preference margins and stabilize iterative improvement. Importantly, local search is only used to construct stronger preference pairs during training and is never invoked at inference time. On TSP and CVRP, ECO achieves the strongest overall performance among the compared neural baselines while also delivering clear advantages in memory usage and throughput. We provide additional analysis on memory scaling, throughput, and the contribution of each design component.

arXiv

A Systematic Investigation of RL-Jailbreaking in LLMs

Montaser Mohammedalamen, Kevin Roice, Reginald McLean, Alyssa Lefaivre Škopac

The evolution of generative models from next-token predictors to autonomous engines of complex systems necessitates rigorous safety hardening. Adversarial jailbreaking, the strategic manipulation of models to elicit harmful output, remains a primary threat to safe deployment. While Reinforcement Learning (RL) frames jailbreaking as a multi-step attack through sequential optimization, a mechanistic understanding of why the framework succeeds remains incomplete. To fill this gap, we present the first systematic decomposition of RL jailbreaking. We deconstruct the framework into problem formalization (reward function, action space, episode length), and algorithmic measures (RL algorithm, training data, reward-shaping) to identify the structural determinants of adversarial success. Our results reveal that the RL-jailbreaker successfully compromised all targeted models and safeguards. Through this first-of-its-kind analysis, we demonstrate that environment formalization, specifically dense rewards and extended episode lengths, is the primary driver of jailbreaking success. This work provides a tool for improving RL-jailbreaker efficiency and, ultimately, harden generative models resistant to RL-based attacks.

Keep browsing

More papers from top AI conferences

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