paperlist.ai

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

The Granularity Gap: A Multi-Dimensional Cross-Generational Audit of Sycophancy in Gemini Models

Patrick Keough

Pass/fail safety evaluation reports whether a model refused. It does not report how far a model went to please the user, and we show these are close to different measurements. We audited sycophancy across three Gemini generations, scoring N=8,830 responses from 8 model variants on 350 adversarial prompts in 7 categories under 3 guardrail conditions, on continuous 1-5 scales for sycophancy, truthfulness and refusal. The judge's own refuse-or-comply verdict explains 29% of the variance in its own sycophancy scores. We term the remainder the Granularity Gap, and it does not close under recalibration: the cut point already in use is the best available on the refusal axis, and no function of that axis explains more than 35%. Reading what four judges wrote while scoring shows why. On a quarter to a third of votes they record that the prompt asked for nothing harmful, almost never in the two categories that solicit a harmful act and up to half the time in the five that do not. A verdict built on refusal has nothing to grade there. Three findings follow. Sycophancy co-occurs with degraded judged truthfulness (rho=0.40), a coupling that strengthens across generations. Capability moved and resistance did not: Gemini 2.0 Flash scores 1.43 and Gemini 3.0 Pro Preview 1.42, with a sharp Gen 2.5 regression between them. And a single direct instruction outperforms an elaborate reasoning protocol in seven of eight variants, cutting mean severity in the most vulnerable category by 60.9%. We evaluate one judge's verdict, not a deployed safety classifier. We release the prompt set, the rubric, and 10,792 per-vote judge scores with their written reasoning.

arXiv

ChartFI: Benchmarking Faithfulness and Insightfulness of Chart Descriptions from Multimodal Large Language Models

Fen Wang, Zekai Shao, Qiman Kang, Chunran Hu, Zhixuan Zhang, Lexu Xie, Chao Liu, Siming Chen

Chart descriptions are essential for accessibility, cross-modal retrieval, and assisting readers in extracting insights from complex visualizations. As multimodal large language models (MLLMs) are increasingly adopted for automated chart description generation, a critical question arises: how faithfully and insightfully do these models actually describe charts? Current benchmarks fall short on two fronts: existing datasets consist of simple, homogeneous charts paired with shallow, fact-enumerating descriptions; and prevailing metrics fail to capture the multi-faceted nature of description quality. To address these gaps, we present the Chart Faithfulness and Insightfulness Benchmark (ChartFI-Bench). We first summarize four dimensions that characterize high-quality chart descriptions: factual accuracy, salient feature emphasis, domain-informed guidance, and chart-text complementarity. Guided by these dimensions, we construct a high-quality benchmark comprising 896 chart-description pairs, which feature visually complex charts and semantically rich descriptions. Furthermore, we design four aligned evaluation metrics -- Faithfulness, Coverage, Informativeness, and Acuity -- to systematically assess the quality of descriptions across these dimensions. Experiments conducted on mainstream MLLMs demonstrate the effectiveness of the proposed framework and reveal common weaknesses among existing models.

arXiv

Hypergraph as Language

Mengqi Lei, Guohuan Xie, Shihui Ying, Shaoyi Du, Jun-Hai Yong, Chuan Shi, Ling Tian, Siqi Li, Yue Gao

Large language models (LLMs) have recently shown strong potential in modeling relational structures. However, existing approaches remain fundamentally graph-centric: they focus on processing pairwise graph structures into tokens that LLMs can understand. In contrast, many real-world relational patterns do not naturally conform to the pairwise-edge assumption, and are better modeled as high-order associations in hypergraphs. For hypergraph structures, existing methods often fail to preserve the native semantics that multiple objects are jointly connected by the same high-order relation, limiting their ability to exploit complex structures. To address this limitation, we put forth the "Hypergraph as Language" perspective and propose Hyper-Align, a hypergraph-native alignment framework for large language models. Hyper-Align compiles the query-object-centered hypergraph context into hypergraph tokens directly consumable by a base LLM. Specifically, we introduce Hypergraph Incidence Detail Template with Overview (HIDT-O), which serializes high-order association structures into a fixed-shape hybrid template combining local incidence details and overview-level summaries. We then design a Hypergraph Incidence Projector (HIP), which maps native high-order incidence structures into the LLM token space through explicit semantic-structural decoupling and bidirectional message passing between vertices and hyperedges. We further define a concrete Hypergraph-as-Language input protocol, which jointly feeds hypergraph tokens and textual prompts into a frozen base LLM, supporting both vertex-level and hyperedge-level tasks under a unified question-answering paradigm. To systematically evaluate different methods in hypergraph structural modeling, we introduce HyperAlign-Bench. Extensive experiments show that Hyper-Align significantly outperforms existing methods across in-domain and zero-shot evaluations.

arXiv

Enhancing Science Classroom Discourse Analysis through Joint Multi-Task Learning for Reasoning-Component Classification

Jiho Noh, Mukhesh Raghava Katragadda, Raymond Carl, Soon Lee

Analyzing the reasoning patterns of students in science classrooms is critical for understanding knowledge construction mechanism and improving instructional practice to maximize cognitive engagement, yet manual coding of classroom discourse at scale remains prohibitively labor-intensive. We present an automated discourse analysis system (ADAS) that jointly classifies teacher and student utterances along two complementary dimensions: Utterance Type and Reasoning Component derived from our prior CDAT framework. To address severe label imbalance among minority classes, we (1) stratify-resplit the annotated corpus, (2) apply LLM-based synthetic data augmentation targeting minority classes, and (3) train a dual-probe head RoBERTa-base classifier. A zero-shot GPT-5.4 baseline achieves macro-F1 of 0.467 on UT and 0.476 on RC, establishing meaningful upper bounds for prompt-only approaches motivating fine-tuning. Beyond classification, we conduct discourse pattern analyses including UTxRC co-occurrence profiling, Cognitive Complexity Index (CCI) computation per session, lag-sequential analysis, and IRF chain analysis, revealing that teacher Feedback-with-Question (Fq) moves are the most consistent antecedents of student inferential reasoning (SR-I). Our results demonstrate that LLM-based augmentation meaningfully improves UT minority-class recognition, and that the structural simplicity of the RC task makes it tractable even for lexical baselines.

arXiv

Faster, Cheaper, More Accurate: Specialised Knowledge Tracing Models Outperform LLMs

Prarthana Bhattacharyya, Joshua Mitton, Ralph Abboud, Simon Woodhead

Predicting future student responses to questions is particularly valuable for educational learning platforms where it enables effective interventions. One of the key approaches to do this has been through the use of knowledge tracing (KT) models. These are small, domain-specific, temporal models trained on student question-response data. KT models are optimised for high accuracy on specific educational domains and have fast inference and scalable deployments. The rise of Large Language Models (LLMs) motivates us to ask the following questions: (1) How well can LLMs perform at predicting students' future responses to questions? (2) Are LLMs scalable for this domain? (3) How do LLMs compare to KT models on this domain-specific task? In this paper, we compare multiple LLMs and KT models across predictive performance, deployment cost, and inference speed to answer the above questions. We show that KT models outperform LLMs with respect to accuracy and F1 scores on this domain-specific task. Further, we demonstrate that LLMs are orders of magnitude slower than KT models and cost orders of magnitude more to deploy. This highlights the importance of domain-specific models for education prediction tasks and the fact that current closed source LLMs should not be used as a universal solution for all tasks.

arXiv

Misconception Diagnosis From Student-Tutor Dialogue: Generate, Retrieve, Rerank

Joshua Mitton, Prarthana Bhattacharyya, Digory Smith, Thomas Christie, Ralph Abboud, Simon Woodhead

Timely and accurate identification of student misconceptions is key to improving learning outcomes and pre-empting the compounding of student errors. However, this task is highly dependent on the effort and intuition of the teacher. In this work, we present a novel approach for detecting misconceptions from student-tutor dialogues using large language models (LLMs). First, we use a fine-tuned LLM to generate plausible misconceptions, and then retrieve the most promising candidates among these using embedding similarity with the input dialogue. These candidates are then assessed and re-ranked by another fine-tuned LLM to improve misconception relevance. Empirically, we evaluate our system on real dialogues from an educational tutoring platform. We consider multiple base LLM models including LLaMA, Qwen and Claude on zero-shot and fine-tuned settings. We find that our approach improves predictive performance over baseline models and that fine-tuning improves both generated misconception quality and can outperform larger closed-source models. Finally, we conduct ablation studies to both validate the importance of our generation and reranking steps on misconception generation quality.

arXiv

Knowing When to Defer: Selective Prediction for Responsible Knowledge Tracing

Joshua Mitton, Prarthana Bhattacharyya, Ralph Abboud, Simon Woodhead

Research on Knowledge Tracing (KT) models traditionally focuses on improving predictive accuracy. However, responsible real-world deployment requires models to know when to defer uncertain predictions to a human teacher. We introduce an intrinsic selective prediction layer for existing KT models using Monte Carlo Dropout (MC-Dropout) to quantify uncertainty. We evaluate this approach across three architectures (DKT, SAKT, and AKT) using the Eedi mathematics dataset. Abstaining on the 20\% most uncertain predictions lifts accuracy by 2.3 to 3.0 percentage points, AUC by 1.9 to 2.4 percentage points and F1 by 1.4 to 4.3 percentage points without any retraining. This abstention strategy is highly targeted: the deferred set exhibits 1.45 to 1.60 times the error rate of the kept set. Furthermore, this targeting holds within every question-difficulty quartile and remains fair across student-ability levels. Importantly, MC-Dropout variance gives roughly five times the AUC lift of a calibrated two-parameter logistic (2PL) Item Response Theory (IRT) baseline as a selective-prediction signal. A variance decomposition of the model's epistemic uncertainty (BALD) reveals that the entire classical psychometric stack, comprising question difficulty, student ability, IRT-style outcome ambiguity, and historical curriculum coverage, explains less than 4\% of the signal under linear modeling and at most 23\% even with a non-linear regressor. This leaves 77\% to 90\% as architecture-specific epistemic content that MC-Dropout surfaces and simpler proxies cannot recover. Selective prediction with model-native epistemic uncertainty is therefore a necessary component of responsible KT deployment, complementary to subgroup-fairness audits and downstream classroom evaluation rather than a substitute for them.

arXiv

A validity-guided workflow for robust large language model research in psychology

Zhicheng Lin

Large language models (LLMs) are rapidly being integrated into psychological and behavioral research as research tools, evaluation targets, human simulators, and cognitive models. Yet recent evidence reveals severe measurement unreliability: personality assessments degenerate under factor analysis, moral preferences reverse with punctuation changes, and theory-of-mind accuracy varies widely with trivial rephrasing. These "measurement phantoms"--statistical artifacts masquerading as psychological phenomena--threaten the validity of a growing body of research. Guided by the dual-validity framework that integrates psychometrics with causal inference, we present a six-stage workflow that scales validity requirements to research ambition--using LLMs to code text requires basic reliability and accuracy, whereas claims about psychological properties demand comprehensive construct validation. Researchers must (1) explicitly define their research goal and corresponding validity requirements, (2) develop and validate computational instruments through psychometric testing, (3) design experiments that control for computational confounds, (4) execute protocols transparently, (5) analyze data with methods appropriate for non-independent observations, and (6) report findings within boundaries and use results to refine theory. We illustrate the workflow through an example of model evaluation--"LLM selfhood"--showing how systematic validation can distinguish genuine computational phenomena from measurement artifacts. By establishing validated computational instruments and transparent practices, this workflow provides a path toward building a robust empirical foundation for AI psychology research.

arXiv

From Prompts to Constructs: A Dual-Validity Framework for Large Language Model Research in Psychology

Zhicheng Lin

Large language models (LLMs) are entering psychological research both as tools and as objects of inquiry. Yet many studies apply human instruments to LLMs without establishing that the outputs are reliable or interpretable, raising the risk of measurement phantoms--statistical regularities mistaken for genuine psychological phenomena. This review argues that robust AI psychological research requires integrating two methodological traditions: psychometric validation of what a score means and causal inference standards for what the results warrant. It develops a dual-validity framework in which evidentiary demands scale with scientific ambition: from tool use through behavioral characterization and human simulation to cognitive modeling. Classifying text may require only accuracy and reliability; claiming that an LLM simulates anxiety or illuminates cognitive mechanisms requires additional evidence, including construct validity evidence and experimental controls. Progress depends on developing computational analogs of psychological constructs rather than assuming human measures automatically apply to language models.

arXiv

The Political Ideology of Large Language Models: Measurement, Inconsistency, and Persuasive Influence

Nouar Aldahoul, Hazem Ibrahim, Aaron R. Kaufman, Talal Rahwan, Yasir Zaki

Large Language Models (LLMs) are a transformational technology, fundamentally changing how people obtain information and interact with the world. As people become increasingly reliant on them for an enormous variety of tasks, a body of academic research has developed to examine these models for inherent biases, especially political biases, often finding them small. We challenge this prevailing wisdom. First, by comparing 43 LLMs to legislators, judges, and a nationally representative sample of U.S. voters, we show that LLMs' apparently moderate overall partisan positioning is the net result of offsetting strongly partisan expressed positions on specific topics, much like moderate voters. Second, in a pre-registered randomized experiment, we show that LLMs can exert persuasive influence on political attitudes. Voters randomized to discuss a policy issue with an LLM shift toward that model's measured ideological position by 3.5 percentage points on average, an effect at least as large as those produced by professional campaign advertising. Explicitly prompting a model to argue one side of the issue shifts attitudes by more than 10 percentage points relative to unsteered conversations, and this steering accounts for the pooled effect. When the same models converse naturally, without steering, we detect no persuasive effect, and our confidence interval rules out effects as small as the pre-registered smallest effect of interest. Contrary to expectations, these persuasive effects are not moderated by familiarity with LLMs, news consumption, or interest in politics. LLMs, especially those controlled by private companies or governments, may become a powerful and targeted vector for political influence.

arXiv

Handover of In-Context Learning State Across Session Boundaries

Masahiro Kato, Taka Kato

This study investigates the methodological and theoretical properties of session handover in applications that use large language models. A task may continue in a new session when the context reaches the model's input limit, when the application restarts, or when another agent is asked to finish the task. The application must then decide which information from the earlier session to pass on. We formulate handover as the transfer of a task-relative in-context learning (ICL) state and distinguish exact recovery of earlier material from preservation of the target distribution. Under an exogeneity condition, predictive equivalence characterizes the coarsest deterministic sufficient handover and gives a fixed-length bit requirement. The analysis isolates the effects of the memory constraint, the writer, and the continuation procedure, and quantifies the cost of writing before the realized downstream query is known. We propose a three-part record that stores decisions and constraints exactly, uses task-justified statistics for repeated evidence, and retains original observations whose effect is not preserved by those statistics. Gaussian linear regression gives an exact finite-dimensional handover and finite-bit perturbation bounds, while nonparametric regression gives upper and lower bounds that relate memory to squared prediction error. These results provide a theory and method for deciding what a handover must retain and how its memory requirement depends on the continuation task.

arXiv

Participatory Moral AI Is Not Neutral: The Invisible Hand of Developers

Taenyun Kim, Edyta Bogucka, Daniele Quercia

As AI systems make more morally loaded decisions across society, one response has been moral preference elicitation. In this approach, researchers poll participants on hypothetical dilemmas and use the aggregated votes to train a policy that an AI model then applies at scale. Before any vote is cast, developers make three key choices in the moral AI elicitation pipeline: feature scoping, voter sampling, and question framing. In other words, they decide which features go to a vote, which voters to include, and how to present the question. These choices are often opaque, undocumented, and treated as technical details rather than normative ones. We examine each of these choices within a common empirical study and show that each can shape the preferences produced by moral AI elicitation. Across two phases (N = 809) in three deployment contexts (i.e., AI kidney allocation, AI agents simulating absent workers, and generative AI depictions of the deceased), we examine the three main stages of the moral AI elicitation pipeline. First, morally relevant features shift across contexts. This suggests that feature schemas should not be assumed to transfer across deployment domains. Second, preferences differ by political ideology for roughly one-third of features, with some differences reversing direction. The ideological composition of the voter pool can therefore affect the resulting aggregated preference profile. Third, the wording of the elicitation question can narrow or widen ideological gaps by up to a full scale point. The framing conditions also change how moral foundations are associated with participants' judgments. Taken together, these findings suggest that voting-based alignment cannot deliver fair or transparent AI by aggregation alone; at minimum, each stage of the moral AI elicitation pipeline should be audited and disclosed.

arXiv

Learning-to-Transition for Large-scale and High-Order MIMO Detection

Yubo Zhang, Yiyao Liu, Xiaodong Wang

High-order multiple-input multiple-output (MIMO) detection requires efficient search over a large discrete symbol space while producing reliable soft information for channel decoding. This paper develops a learning-to-transition (L2T) framework that formulates MIMO detection as a stochastic sequence of complete-vector transitions. At each transition, a channel-coupled Transformer updates both the instance embedding and the sampling policy, while a blockwise autoregressive factorization captures inter-stream dependence with moderate sequential complexity. For hard-output detection, a transition network is applied recursively and trained through a residual-to-BER curriculum, which first learns the MIMO search geometry from the exact residual metric and then aligns the policy with transmitted-bit accuracy. For soft-output reception, the well-trained hard policy is cloned at the parameter level into every layer of an untied soft-input soft-output iterative detection and decoding (IDD) receiver. This tied-to-untied transfer preserves the learned zero-prior search dynamics while enabling layer- and round-specific specialization under decoder feedback. Within each IDD round, decoder priors tilt candidate generation according to Bayes' rule, and likelihood-weighted terminal hypotheses produce posterior and extrinsic log-likelihood ratios for LDPC decoding. A multi-stage training strategy further stabilizes the hard-to-soft transfer by progressively exposing the receiver to synthetic and in-loop decoder-generated priors.

arXiv

Optimal Scheduling of Road Maintenance Jobs Considering Impact on Traffic Flows

Charitha Nandepu, Lohitha Kalepu, Gabriele Ciavarella, Sangwoo Park

Network-level maintenance planning requires repeated evaluations of equilibrium traffic flows under road capacity reductions. While equilibrium traffic assignment models are well established, their repeated solution quickly becomes computationally prohibitive and challenging to embed within maintenance scheduling problems. This paper investigates data-driven surrogate models that approximate equilibrium arc flows directly from origin-destination demand, using optimization-based equilibrium solutions as ground truth. A real-world case study based on traffic data from the Newark, New Jersey area demonstrates the effectiveness of the proposed approach as a scalable building block for future maintenance scheduling frameworks.

arXiv

Twin: Playing an Unknown Game with a Test-Time Digital Twin

Alexy Skoutnev, Kirill Acharya, Gaston Longhitano, Madeleine Udell, Kevin Ellis, Iddo Drori

We present a Test-time World-model Inference (Twin) system, in which a frontier coding agent writes an executable world model for completing continual learning tasks, such as ARC-AGI-3 games. Traditional approaches hand-engineer such models, one custom design per task. Each game hides its rules and goal, and our system constructs them from simulation and interaction alone. Its inductive prior over grid games is strong enough to recover the true transitions of the game and the goal on nearly all levels. Replay validation happens in a twin world model. The harness enforces that an action is not made until the program reproduces every previous observed game transition. Each mismatch between a world model prediction and the actual action result becomes a counterexample that is used to repair the world model. Twin clears 179 out of 183 levels (97.8%), and does so more efficiently than humans in 158 out of 179 levels (88.3%). The system infers the goal before any reward on 156 of the levels it clears (87.2%), and in the remaining levels automatically discovers the goal by search. The benchmark scores completion and action efficiency, between 0 and 100, against humans playing each game for the first time. Played directly, the base model scores only 7.8%; an off-the-shelf harness increases it to 61.1%, whereas our twin world model increases the same base model to 93.3%, clearing 23 out of 25 games. Building a usable world model is simpler than anticipated, whereas the harder problem is inferring the right goal.

arXiv

Ensuring Safe Physical AI in Urban Mobility via Hazard-Informed Synthesized Envelopes

Alexei Odinokov, Rostislav Yavorskiy

As heterogeneous robotic systems deploy across diverse urban zones, maintaining safety amid complex human-robot interactions remains a critical challenge. We present a unified framework that bridges systematic hazard analysis and runtime enforcement using hazard-informed safety envelopes. Rather than treating safety as a static constraint isolated within individual software modules, we introduce a cross-layer safety transformation process spanning symbolic, spatial, and dynamic world models. We show how this representation naturally interfaces with physical AI runtime harnesses to guarantee safe urban mobility.

arXiv

Shift Aware Transfer Learning with Adaptive Dual-Encoder Fusion for PM Forecasting in Data-Limited Environments

Shahab Band, Hamed Mohammadi

Short-horizon forecasting of fine particulate matter (PM2.5) remains difficult when observations from the target domain are limited and the statistical properties of the source and target domains differ. In these settings, models trained only on local data may not capture complex temporal dynamics, while direct transfer learning can result in negative transfer. This study develops a shift-aware dual-encoder transfer framework that combines source-domain knowledge with target-specific representation learning. The source encoder was pretrained using hourly observations from 10 U.S. monitoring locations. The framework was then adapted and evaluated using two years of hourly observations from 77 stations in Taiwan under a chronological train-validation-test protocol. Among the four principal baselines, the frozen-source dual-encoder model achieved the best performance, with MSE = 21.8960, MAE = 3.1597, and R^2 = 0.8725. This corresponds to an MSE reduction of approximately 7.1% relative to TL-v1 and 4.1% relative to TL-v2. The ablation analysis showed that removing the Taiwan-specific branch caused the largest decline in performance. Allowing the source encoder to adapt produced the best overall result, with MSE = 21.6575, MAE = 3.1383, and R^2 = 0.8739. SHAP analysis indicated that predictions were driven mainly by recent PM2.5 observations and meteorological variables related to pollutant transport and dispersion. These results suggest that source-domain knowledge is most effective when target-specific information is preserved and the transferred representation is allowed to adapt under target supervision.

arXiv

SheetCompass: Hierarchical Relation Graphs for Agentic Spreadsheet Reasoning

Panjing He, Mingyue Cheng, Yucong Luo, Li Li, Xiaohan Zhang

Spreadsheets are widely used to organize, analyze, and manipulate semi-structured data, yet automated spreadsheet reasoning remains challenging for large language models (LLMs). Real-world workbooks often contain implicit cross-table associations, fine-grained column dependencies, and complex spatial layouts. Existing methods typically flatten these multidimensional structures into sequential strings, losing important intra-sheet boundaries and inter-sheet semantics. Consequently, LLMs cannot exploit the global spatial context that human experts naturally use when inspecting spreadsheets. We propose SheetCompass, a graph-guided and memory-driven agentic framework for spreadsheet reasoning and automation. SheetCompass explicitly models structural relationships within and across worksheets while maintaining task-relevant information in memory, enabling agents to reason more effectively over complex workbooks.

arXiv

Wyvern: An Agentic Framework for Generating Grounded Multimodal Reports

Beatrice Alessandra Motetti, Emilien Guandalino, Daniele Jahier Pagliari, Alessio Burrello, Lorenz K. Müller, Konstantin Berestizshevsky, Lukas Cavigelli

In the current artificial intelligence-driven innovation era, the pace of knowledge growth is accelerating, and is hard to keep up with. While generative models are increasingly used to synthesize content, they often lack in information grounding. To address these peculiarities of our time, we propose Wyvern, a multi-agent framework for the automated generation of grounded, multimodal technical reports. Wyvern allows for the generation of multimodal outputs, integrating images, tables, and text with supporting references in a unified report. Additionally, a particular focus is placed on the grounding of the content, with the implementation of a claims auto-revision stage. We conduct a human evaluation study to assess the quality of our proposed framework. The results show that the figures' informativeness is perceived as superior to that of a recent baseline in 87% of cases. Furthermore, Wyvern's reports are rated as more useful than those produced by three alternative methods in 63% to 100% of instances. We also carry out automatic evaluations showing that Wyvern gains up to 2.3$\times$ in citation recall and 1.6$\times$ in citation precision with respect to the baselines.

arXiv

PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments

Yuhao Zhan, Bingxiang He, Zecong Tang, Chaojun Xiao

Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change. To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simulator-grounded benchmark of 144 source-to-target adaptation pairs across six physics domains. Each pair links a source environment to a mutated target environment with the same goal and interface. A code-driven design that succeeds in the source fails in the target, where agents must iteratively adapt it into a working target design using diagnostic sandbox feedback within a limited attempt budget. We compare ten self-evolving methods from four paradigms. The benchmark remains far from saturated: Reflexion + Qwen3-14B succeeds on only 35.9\% of full-benchmark pairs, while GPT-5.5 solves 66.7\% of the Statics subset under the full budget. Together, these results show that simulator-grounded reflection is more reliable than unverified self-revision, while memory anchors agents to early designs and broad tree search explores without converging. Even revealing exact physical changes does not raise the performance ceiling, pointing to mechanism redesign rather than parameter inference as the central bottleneck. Data and code are available at https://github.com/thunlp/PACE-Bench.

Keep browsing

More papers from top AI conferences

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