Artificial Intelligence
A Generalization of Amari's Bayesian Duality
Mohammad Emtiyaz Khan and Thomas Möllenhoff
Amari's contributions to information geometry and machine learning are well known. Here, we revisit Amari's work on
Bayesian duality which has not received as much
attention. We connect Amari's
Bayesian duality to a convex duality of Bayes' rule. Using this connection, we present a generalization of Amari's
Bayesian duality and discuss its relevance for modern artificial intelligence.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CV
Canonical Color as a Lens into Concept Decodability in Vision Encoders and VLMs
Xiaofu Chen, Stella Frank and Yova Kementchedjhieva
Visual encoders construct a representation of the image input for
Vision-
Language models. How much conceptual, as opposed to immediately visible, information does this representation contain? We use canonical color as a controlled test case to ask whether
vision encoders make canonical-color information linearly accessible, even when color is removed from the input image. We construct a
dataset of objects with canonical colors, and probe
vision encoders for both color and object identity using color and grayscale images. We find that canonical color remains decodable from grayscale images, and is tied to predicted object identity, indicating a conceptual link. Extending this analysis to full VLMs, we find that VLM post-training can have a surprisingly large effect on color decodability in the
vision encoder. Overall, canonical color provides a usefully controllable lens for tracing object-level conceptual semantic information in
vision encoders and VLMs.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.RO
DeCAL: Towards Physically-Grounded Dexterous Vision-Language-Action Models via Contact-Aware Latent Co-Imagination
Yankai Fu, Ning Chen, Junkai Zhao et al. (+5 more)
Dexterous manipulation involves contact-rich and fine-grained interactions with the physical world, posing significant challenges for existing
vision-
language-action (VLA) models due to severe visual occlusions and complex contact dynamics. While recent works have incorporated tactile sensing into robotic manipulation, most approaches still rely on homogeneous
multimodal fusion, lacking adaptive tactile integration and explicit modeling of physical dynamics. In this work, we present DeCAL, a physically-grounded dexterous
vision-
language-action model that unifies understanding, imagination and action generation for contact-rich dexterous manipulation. Built upon a Mixture-of-
Transformers (MoT) architecture, DeCAL leverages specialized experts for each capability while enabling efficient information flow among them. To effectively leverage tactile information, we introduce Adaptive Visuo-Tactile Fusion that dynamically regulates tactile interactions via a contact-aware gating strategy. Furthermore, we propose Visuo-Tactile Latent Co-Imagination to jointly model visual and tactile dynamics, equipping the policy with implicit physical world knowledge. Experimental results show that DeCAL consistently achieves state-of-the-art performance across all tasks, attaining a 71% average success rate and an 83.4% progress success rate, while also demonstrating strong generalization to unseen scenarios. The website is available at https://aureleopku.github.io/DeCAL.
Read the full abstract ▾
All authors Yankai Fu, Ning Chen, Junkai Zhao, Heng Zhang, Guocai Yao, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang
Artificial Intelligence
MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents
Boyu Yang, Jiazheng Sun, Zilong Lu et al. (+3 more)
Long horizon Large
Language Model (LLM)
agents rely on external
memory systems to preserve user preferences and task knowledge across extended interactions. Conventional
retrieval mechanisms optimize semantic compatibility rather than downstream utility, frequently introducing outdated, misleading, or conflicting evidence into the active context. We present MeClear, a task conditioned
memory clearance framework that identifies
memories featuring negative downstream utility through cooperative attribution and selectively suppresses them from
agent execution. MeClear combines Leave One Out screening with sampled cooperative Shapley attribution to distribute utility across interacting evidence, effectively resolving redundant conflict masking where single removal evaluations fail. Utilizing attribution rankings, MeClear executes a query scoped minimal clearance strategy over a nested filtration, verifying task recovery on the cleared context without permanently altering the persistent
memory bank. Comprehensive experimental evaluations across ten long dialogue
memory pools demonstrate that MeClear achieves a target recall of 85.9% and an overall task recovery rate of 82.3%, representing a 25.5 percentage point improvement over Leave One Out (LOO) baselines.
Read the full abstract ▾
All authors Boyu Yang, Jiazheng Sun, Zilong Lu, Zhi Qiu, Xin Peng, Jun Zheng
Artificial Intelligence
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
Yuqiao Tan, Shizhu He, Jun Zhao et al. (+1 more)
While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and
auditing to understand what models learn and ensure safe
alignment. Mechanistic interpretability tools are essential to bridge this gap, among which Sparse Autoencoders (SAEs) serve as a cornerstone by isolating interpretable features for model inspection and steering. In this paper, we introduce SAEScientist-Bench to evaluate whether AI
agents can act as scientists utilizing SAE tools for autonomous mechanistic discovery. Given a target concept, an
agent designs contrastive probes and navigates a Gemma Scope dictionary of 131K+ features in Gemma-2-9B-IT to discover the optimal feature, evaluated against curated expert reference features anchored on Neuronpedia across activation rank, concept selectivity on contrastive texts, and causal steering. Across 10
agent configurations and 20 tasks, frontier
agents demonstrate genuine discovery capabilities and lead different evaluation dimensions, but remain well behind the expert baseline, approaching expert levels on separating target concepts from contrastive controls while lagging substantially in causal generation steering. Further analysis reveals that although
agents can design contrasts to rule out spurious candidates, they frequently misinterpret experimental measurements. These results establish experimental model understanding as a measurable capability for closed-loop autonomous AI R&D. Our code is available at https://github.com/Trae1ounG/SAEScientist.
Read the full abstract ▾
All authors Yuqiao Tan, Shizhu He, Jun Zhao, Kang Liu
Artificial Intelligence
The Surprising Effectiveness of Approximate Value Iteration in Self-Play
Raphael Boige, Amine Boumaza and Bruno Scherrer
Combining search with function approximation has driven major advances in game-playing programs, making self-play
algorithms more competitive than ever. Still, the computational overhead of the most popular methods, based on Monte Carlo Tree Search (MCTS), can be substantial. In this work, we investigate whether simpler methods remain competitive in non-trivial, moderately sized games such as Connect Four, Hex(7x7) and synthetic games. We train a minimal self-play implementation of Approximate Value Iteration (AVI) and use ground-truth oracles for exact evaluation. Contrary to expectations, our results demonstrate the surprising effectiveness of AVI: it learns more accurate value functions than those learned by AlphaZero, while its one-step-lookahead greedy policies remain competitive with MCTS-based policies at substantially lower training and inference costs. Preliminary experiments on Othello and Go(9x9) show that AVI trains stably on larger games and learns effective value functions. These findings suggest that the success of MCTS-based methods may have eclipsed simpler approaches that have become increasingly practical with modern deep-learning tools.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CL
SPINE: Measuring LLM Sycophancy under Sustained Multi-Turn Pressure
Leyuan Tang, Kangda Wei, Tianyu Jiang et al. (+1 more)
Large
language models (LLMs) may abandon correct positions when users push back, exhibiting a failure mode known as sycophancy. Existing evaluations typically use short, prespecified conversations and may therefore miss failures that emerge under sustained, adaptive disagreement. We introduce SPINE, a
benchmark in which an LLM proxy plays a persistent but mistaken user and adaptively challenges a target model for up to 25 turns. We evaluate four commercial models (Claude Sonnet 5, GPT-5.6 Terra, Gemini 3.1 Pro, and DeepSeek V4 Pro) and three Olmo3-7b models on 100 false presuppositions and 100 unethical queries. We define collapse as a target model abandoning the correct position and yielding to user pressure. Our experimental results show that collapse rates increase with conversation length across post-trained models, with the overall rate rising from 0%-57% within 5 turns to 20%-97% within 25 turns depending on model and scenario. These results show that evaluations limited to a few conversational turns underestimate sycophancy and resistance under sustained pressure remains unreliable across current models. By analyzing models with accessible
reasoning traces, we surprisingly found that the correct position often remains represented in the trace in 50%-89% of analyzed collapse turns, suggesting that the model chooses to please a user and sycophancy is not due to ignorance or the correct knowledge being overridden. Ablations show that the SPINE protocol substantially increases the sycophantic collapse rate compared to pre-generated scripts. Among all tactics, emotional appeals are most associated with inducing LLM sycophantic behavior, with position weakening in 44.3% of turns using this tactic. Code, data, and transcripts will be released upon publication.
Read the full abstract ▾
All authors Leyuan Tang, Kangda Wei, Tianyu Jiang, Ruihong Huang
Artificial Intelligence · Cross-listed from cs.CV
GoDeep: Annotation-Free Open-Vocabulary 3D Scene Understanding via Language-Space Lifting
Thodoris Betsas, Anastasios Doulamis and Andreas Georgopoulos
Open vocabulary 3D semantic
segmentation methods typically lift CLIP features into 3D. This embeds points in a joint
vision-
language space known to behave like a bag-of-words on compositional tasks. Furthermore, even annotation free variants often require a large 3D training corpus and a dedicated 3D encoder per domain. Instead we use a
vision-
language model purely as a translator. It produces structured, entity-level descriptions of each posed image. These descriptions are grounded, projected, and aggregated directly in a general-purpose,
language-only
embedding space, with no 3D training corpus or encoder required. On ScanNet++, our pipeline is competitive with strong annotation free baselines trained on ScanNet. On a 5-building cultural heritage
benchmark, raw scores initially favor a CLIP-based variant, but a single systematic vocabulary correction reverses this ranking. An effect confirmed by a second, independent correction on a different class, indicating that
language-space
embeddings track physical content more faithfully. This fidelity extends to genuinely out-of-vocabulary (OOV) objects on ScanNet++ proving that
language-space
embeddings separate presence from absence objects far more sharply than CLIP-based
embeddings do. GoDeep also localize these OOV objects within the scene, all without any 2D-3D annotation. Because every representation remains discrete text, predictions are also explainable at the point level. Finally, exploiting both a heuristic weighting, that favors precise over merely frequent observations and GoDeep's explainability property, we propose an aggregation strategy, as a proof of concept, that favors finer elements localization.
Read the full abstract ▾
Artificial Intelligence
Everything in Moderation: Per-Domain Coverage Optima and Alignment-Resistant Domain Gaps in Multi-Domain Mid-Training
Yunpeng Xu and Kun Zheng
Mid-training, the stage between pre-training and
alignment, is where a model's per-domain data composition is typically set by data availability rather than principled design. We ask what that decision buys, and whether a later
alignment pass can undo it. In a controlled logical-
reasoning setting (Qwen3-8B-Base, with a 4B replication; five semantically rule-disjoint KOR-Bench domains) we train 30 allocations spanning the five-domain simplex, 24 sweep configurations plus six withheld from the fit, at five seeds each. Three findings emerge. First, every domain has an interior coverage optimum: the moderate band ($10\%$-$40\%$) is best for all five domains, and a calibrated permutation test for quadratic interiority gives $P\approx0.010$; the fitted mid-training-only curves, with 8B peaks between $9.9\%$ and $35.1\%$, reproduce for curve shape but not peak location. Second, the gaps survive a fixed-budget
alignment pass: compensatory SFT raises 116/120
cells (mean $+4.32\%$) yet bridges $0/240$ pairs at a $5\%$ threshold and $30/240$ at a $10\%$ ratio, an equal-budget uniform control behaves almost identically, and a permutation null would bridge $13.8\pm3.3$ and $77.9\pm8.5$ pairs ($P<0.001$). Third, zero coverage collapses mid-training-only accuracy, though a FineWeb-Edu-only control shows the collapse is commingled with generic drift. An exploratory $\theta^*$ allocation attains the largest full-pipeline gain ($+4.36\%$ vs. $+0.80\%$/$+0.64\%$\,pp) but is marginal under Welch test.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
ThinkPrior: Zero-Rollout Difficulty Priors for Cold-Start Prompt Selection in RLVR
Tommy Sha, Skylar Zhai and Siqi Zhao
In reinforcement learning with verifiable rewards (RLVR) trained with group relative policy optimization (GRPO), the KL-free reward-advantage term studied here depends on within-group reward variation. If all rollouts in a group are correct or all are wrong, their group-relative advantages are identically zero; these zero-advantage silent groups provide no reward-advantage gradient, yet uniform sampling spends 39% of a run's rollouts on them. History-based
prompt selection must first spend target-policy rollouts to estimate difficulty, creating a cold start with rollout waste; ThinkPrior instead uses an external anchor in one offline pass to construct a zero-rollout difficulty prior before the first target-policy rollout. The verifier-scored anchor pass rate supplies an external-anchor initialization for a Beta posterior; ThinkPrior selects by expected learnability and then updates from training outcomes, changing neither the loss nor the optimizer. On Qwen2.5-Math-7B across sixteen seeds, ThinkPrior more than halves early silent groups and cuts wasted rollouts through step 30 by nearly a fifth, while we detect no difference in final accuracy. On this 250-
prompt pool the fixed-budget result is a reallocation rather than a net saving. The measured ThinkPrior+DAPO composition reduces generated rollouts by 10.6% while both arms retain the same 3840-rollout update budget. The prior requires no target-policy rollout before the first selection, but the posterior thereafter uses target-policy outcomes.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CL
Performance of Clinical AI System and Physicians and Frontier Language Models in primary care diagnostics
Andy Nkansah, Hanna Plotnitskaya, Stanislau Salavei et al. (+6 more)
Clinical AI evaluation should encompass diagnosis and management after adaptive information gathering. We compared Doctorina, eight physicians and four standalone frontier
language models in 150 synthetic Polish-
language primary-care consultations. Doctorina achieved 82.0% Top-1 concordance versus 57.0% for physicians (difference, 25.0 percentage points; 95% confidence interval, 17.7-32.7) and 97.3% versus 85.0% primary-or-reference-differential concordance. Across 149 case pairs, normalized workup and treatment scores were 89.4 versus 66.9 and 83.7 versus 61.2. Doctorina had the highest diagnostic point estimates among all six groups; Kimi K3 ranked next, while Claude Opus 5 led the closely spaced management estimates of Opus, Doctorina and Kimi. A second Doctorina execution reproduced the advantages over physicians across all outcomes. Doctorina's advantage over physicians therefore extended from primary-diagnosis selection to higher-rated diagnostic workup and initial treatment after adaptive consultation.
Read the full abstract ▾
All authors Andy Nkansah, Hanna Plotnitskaya, Stanislau Salavei, Anna Kozlova, Piotr Gibas, Julian Milek, Viktar Harbachou, Aleksey Ropan, Pavel Satalkin
Artificial Intelligence · Cross-listed from math.CT
Time-Varying Data as Sheaves: an Invitation to Narratives
Wilmer Leal, Benjamin Merlin Bumpus, Jana K. Nickel et al. (+3 more)
Modern science and engineering increasingly rely on time-varying data, yet the mathematical tools used to model temporal phenomena are often developed within separate disciplines, obscuring common principles and limiting the transfer of ideas across fields. This chapter presents the theory of narratives, an abstract framework for time-varying objects of any mathematical kind that supports both theoretical investigations and applications. To illustrate this perspective, the chapter develops three vignettes, each illustrating a different research direction. The first addresses a general concern: What information loss can occur when switching between different representations of temporal data? The second concerns structural and algorithmic approaches: How can we systematically decompose time-varying data into simple pieces and obtain invariants describing its structural complexity? The third is an application to control theory: How can we model multi-
agent systems with switching communication
topologies? More important than any individual vignette, the central message of this invitation is that a suitable abstract perspective can organize and guide research across remarkably diverse mathematical and scientific domains.
Read the full abstract ▾
All authors Wilmer Leal, Benjamin Merlin Bumpus, Jana K. Nickel, Johan García, James Fairbanks, Warren Dixon
Artificial Intelligence · Cross-listed from cs.LG
Training-Free Task Vectors for LLM Behavioral Control
Gabriel J. Perin, Lucas Boscaini, André Araujo et al. (+1 more)
Task vectors enable post-training model editing by identifying semantically meaningful directions in weight space, typically computed as the difference between a fine-tuned model and its pretrained initialization. However, this reliance on fine-tuning makes discovering such directions costly and limits the practicality of post-training model editing. To address this limitation, we introduce Training-Free Task Vectors (TFTVs), a novel method to compute task-vector-like directions without requiring fine-tuning. Our method maps activation steering vectors to rank-one weight-space edits using only forward-pass statistics, while satisfying arithmetic properties that directly support learning via addition, forgetting via subtraction, and the composition of multiple edits. Empirically, we evaluate TFTVs on large
language model behavioral control tasks and show that they consistently amplify, suppress, and compose target behaviors while preserving general knowledge and problem-solving skills. We also validate our method against other editing and steering baselines, experimentally demonstrating that TFTVs achieve stronger trait control with better or competitive utility preservation. We hope our work opens new directions for the community in post-training model editing and broader training-free model control. Code is available on the project website: tftv-llm.github.io.
Read the full abstract ▾
All authors Gabriel J. Perin, Lucas Boscaini, André Araujo, Nina S. T. Hirata
Artificial Intelligence · Cross-listed from cs.CL
The Audit Decides the Verdict: Instrument Effects Rival Demographic Bias in LLM Decision Audits
Siddharth Vohra and Manikandan Ravikiran
Whether a
language model looks demographically biased can depend on how the audit asks its question. A charitable-aid
benchmark reports that the same models favor minority applicants when rating requests one at a time and penalize some when ranking side by side. We test whether that reversal generalizes to hiring, lending, and medical triage: 40,726 requests to five models, applications differing only in the applicant's name, and a primary test fixed before collection. It does not. None of 36 planned contrasts survives correction. The rating advantage keeps its sign at roughly half the published size, and a precision extension bounds any hiring ranking penalty below the published effect, though the lending and triage ranking floors sit above that margin, so the exclusion is conclusive for hiring ranking and for rating in all three domains only. Planted disparities tracking their injected sizes and a directional replication on the original aid materials bound these nulls. The audit is livelier than the demographics: models recognize transparent audits nearly always, tie every identical-content comparison whether the varying detail is race or a hobby, and reward first-listed candidates as much as any demographic effect we measure. Audit verdicts reflect audit construction more than demographic bias.
Read the full abstract ▾
Artificial Intelligence
Answer-Distribution Trajectories: A Stochastic-Dynamics View of LLM Reasoning
Mar Gonzàlez I Català, Haitz Sáez de Ocáriz Borde, Davide Murari et al. (+3 more)
Chain-of-thought
reasoning provides a structured computation between a model's input and final answer. Yet it is often evaluated through endpoint accuracy, which ignores the path taken to reach that answer. An emerging line of work addresses this limitation using
entropy profiles, which track how uncertainty evolves over the
reasoning process but do not reveal which competing hypotheses account for that uncertainty. We introduce answer-distribution trajectories, a
stochastic-dynamics-inspired representation that tracks the model's full predictive distribution over answers as
reasoning unfolds. As a strictly finer representation than endpoint and
entropy summaries, answer-distribution trajectories enable us to characterize a trace through a dynamical
reasoning profile spanning exploration, revision, motion, and commitment, and to distinguish different dynamical mechanisms of
reasoning success and failure. Across sixteen open-weight
language models and four
reasoning benchmarks, we show that traces with the same endpoint and similar
entropy profiles can exhibit substantially different
reasoning dynamics. We further find substantial variation in these dynamics both within and across models and tasks, with different objectives favoring different dynamical profiles. Additionally, we show that training and inference choices systematically reshape these profiles. Our results suggest that answer-distribution trajectories provide a rich framework for analysing and evaluating the dynamics of LLM
reasoning.
Read the full abstract ▾
All authors Mar Gonzàlez I Català, Haitz Sáez de Ocáriz Borde, Davide Murari, Carola-Bibiane Schönlieb, Pietro Liò, George Montañez
Artificial Intelligence · Cross-listed from cs.SE
It Is Not My Code Anymore
Augusto Camargo
AI-assisted programming raises distinct questions about who produces code, who feels ownership of it, and who is responsible when it fails. This research note examines these distinctions through a hypothetical enrollment failure and a selective reading of the literature. Identifying the producer of a defective expression does not, by itself, determine the duties of reviewers, release decision-makers, or service operators. Collective ownership likewise leaves those duties to be specified. The discussion then considers how quality engineering can evaluate both generated implementations and the processes that produce them. Acceptance criteria should be justified by the required service outcome, with component checks contributing evidence toward that outcome. This perspective also permits comparison with systems that perform a task without generating a separate program for it. Such substitution would change the object of authorship while leaving the service obligation intact. The note reports no new empirical results; it proposes distinctions and evaluation questions for AI-assisted software production.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Let It Go or Learn to Self-Correct: Continuous Diffusion for Constrained Discrete Tasks
Mariia Drozdova, Stéphane Liem Nguyen and François Fleuret
Denoising
Diffusion Probabilistic Models (DDPMs) generate samples by starting from noise and repeatedly denoising while keeping each update close to the current noisy state. This behavior is effective in many continuous domains, but its role is less clear for globally constrained discrete tasks, such as Sudoku, graph connectivity, Latin squares, and N-queens. In such settings, early discrete errors can be difficult to undo. As a result, standard
diffusion sampling may preserve early mistakes, even when the model's clean predictions are informative. We compare standard samplers to sampling directly from the model's clean prediction. Without retraining, this single change improves Sudoku validity from 31% to 95%, with consistent gains across the other discrete tasks. We hypothesize that staying close to the current noisy state is harmful because the reverse trajectory can drift off the forward noising distribution the model was trained on. To reduce this train-test mismatch, we further introduce self-correction training, which exposes the model to its own predictions, improving robustness to errors that arise during inference. This substantially improves the performance of standard samplers. Our results suggest that continuous
diffusion models can learn nontrivial global constraints, but discrete
reasoning tasks require better
alignment between training and inference: either through samplers that reduce commitment to early decisions, or through training that teaches the model to correct its own inference-time errors.
Read the full abstract ▾
Artificial Intelligence
Deposon: An Auditable, Conservation-Guaranteed, Game-Theoretically Tested Scattering Layer over LLM Reasoning Paths
Qihao Yuan
Multi-step LLM
reasoning lacks a machine-recheckable ledger: discarded
reasoning paths leave no auditable record. We propose the Deposon scattering layer, which binds each node of an LLM-generated concept-decomposition graph to a two-parameter Deposon state; paths undergo three-channel scattering – transmission, reflection, irreversible dissipation – obeying T+R+A=1 for arbitrary parameters, with a maximum per-path energy-audit deviation of 2.2E-16 (machine epsilon). We report all three evidence tiers honestly. On synthetic trap
benchmarks the path-filtering gain is closed (pre-registered): unified reaches 100% versus a decoy-capture baseline at 7%/10%. On real
benchmarks the layer is indistinguishable from a trivial six-keyword rule filter (GSM8K 0.87 >= 0.85, McNemar p=0.5; StrategyQA 0.899 = 0.899); no difference is detected here, so we sharpen the claim to "the differential value lies solely in machine verifiability." Fusion yields a second negative result: convex combinations with a semantic prior never improve (physics 0.484 -> 0.452), and the apparent lambda=2 gain is an anti-field artifact; any fusion gain must be nonlinear. Modeling the reverse dynamics as a potential game on the graph, we evidence an auditable scalar's monotonicity and near-gradientness and quantify the empirical coordination ratio (ECR). The three formalized dynamical-equivalence propositions (P1a/P1b/T-P1c) are falsified under the pre-registered kill protocol, and the potential-game claim is downgraded to approximate (cyclic-graph median residual 0.669): only consistency-level evidence survives at the dynamical level. Code: github.com/zeroandcat/Deposon.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Transformers as In-Context Samplers: From Closed-Form Diffusion to Estimation-Free Sampling
Arman Adibi, Alireza Jafari, Mohammad Ghavamzadeh et al. (+1 more)
A growing body of work establishes that large
language models are not mere statistical memorizers, but are capable of in-context learning: performing inference at test time using only examples provided in the
prompt, without any parameter updates. Prior theoretical work has shown that this capability extends to supervised learning tasks such as linear regression. We prove that in-context learning extends further to data generation: frozen
transformers can simulate iterative generative samplers from in-context samples. We first show that
transformers can realize closed-form and smoothed closed-form
diffusion samplers. The construction identifies a concrete generative role for softmax
attention: it computes responsibility weights and weighted empirical averages, while feedforward layers implement Euler updates. To empirically relate these constructions to pretrained
language models, we study semantic-topic sampling:
prompts consisting of words drawn from a common semantic category, such as animals, foods, or cities. Across
transformer layers, the normalized hidden states exhibit a two-stage geometry: they move toward a uniform spherical reference in intermediate layers and then return to structured, topic-dependent representations near the output. We further measure an interacting-particle energy on these hidden-state clouds and observe the same U-shape pattern. We then prove that
transformers can approximate an energy-based sampler, constructing the same U-shape energy across the layers.
Read the full abstract ▾
All authors Arman Adibi, Alireza Jafari, Mohammad Ghavamzadeh, Hadi Daneshmand
Artificial Intelligence · Cross-listed from eess.AS
Orantqing, Shengpeng Ji, Junlong Tong et al. (+19 more)
In this work, we present Gander, a native
multimodal duplex interaction model that builds on MiniCPM-o 4.5 and is further adapted for realtime interaction with an asynchronous
agent loop. In contrast to conventional turn based systems, Gander continuously processes streaming user inputs, enabling full-duplex interaction in both everyday conversations and complex workflow
agent scenarios. Users can interrupt an ongoing response, while the model can proactively provide intermediate feedback or ask follow up questions. To natively support these capabilities, Gander adopts two key architectural designs: 1) a Cerebellum-
Brain collaborative framework, Cerebellum is responsible for realtime interaction while the
Brain handles complex
reasoning and higher level agentic tasks. The two components interact continuously through tool calling and the
agent orchestration runtime. 2) The Cerebellum is built upon a streaming Thinker-Talker architecture, where user inputs and model outputs are flattened into an ordered
token stream at the chunk level. We evaluate Gander across conversational ability, interactive capability, understanding, and tool assisted task execution. Internal human evaluations show that Gander maintains natural and expressive spoken dialogue, while
benchmark results demonstrate effective turn taking capability and encouraging results on spoken question answering and related understanding tasks. Gander also supports a range of challenging interaction settings, including background noise interference, multi-party interactions, and backchannel communication. While our current evaluation focuses on tool assisted settings, broader long horizon
agent tasks and more diverse deployment conditions remain promising directions for further study. We release Gander together with its models, code, and data to facilitate further research and development in the community.
Read the full abstract ▾
All authors Orantqing, Shengpeng Ji, Junlong Tong, Jialong Zuo, Dongjie Fu, Di Cao, Yangzhuo Li, Shangda Wu, Franz, Evan, Theron Veyra, Changhao Pan, Jingyu Lu, Dongchao Yang, Zhifei Xie, Yang Tan, Xiaoyu Shen, Xiaoda Yang, Wenfu Wang, Teddy Sun, Steve Yves, Zhou Zhao
Artificial Intelligence · Cross-listed from cs.LG
GraphFAS: A Distributed System for Automated Graph Feature Generation and Selection in Industrial Transaction Networks
Yice Luo, Yun Zhu, Xi Chen et al. (+7 more)
Industrial fraud detection often relies on costly expert-crafted features that overlook graph-structured relational signals, while GNNs often do not meet the interpretability and deployment requirements of financial risk control. We propose GraphFAS (Graph Feature Automated Selection), a distributed feature selection procedure based on Boruta that bridges this gap through: (1) a non-parametric graph feature generation module that constructs explicit, interpretable structural features via multi-hop subgraph extraction and multi-scale aggregation without learned parameters; and (2) an automated distributed feature selection
algorithm extending Boruta with median-based aggregation across partitions to robustly identify informative features at scale with minimal domain expertise. Compared with end-to-end GNN pipelines, GraphFAS decouples feature aggregation from model training, enabling direct integration with tabular models and direct compatibility with TreeSHAPbased explanations. Deployed in Alipay, GraphFAS delivers orderof-magnitude improvements in engineering efficiency while showing strong performance against expert-driven and graph-learning baselines on large-scale graphs.
Read the full abstract ▾
All authors Yice Luo, Yun Zhu, Xi Chen, Yongchao Liu, Xintan Zeng, Chengying Huan, Kai Zhang, Jinrui Zhang, Juelu Zhang, Jiajun Zheng
Artificial Intelligence
Good Pretraining, Bad SFT: Checkpoint Quality Across the Training Stack
Sohir Maskey, Philipp Scholl, Jonas Knupp et al. (+2 more)
Language-model checkpoints are commonly selected by pretraining loss or
benchmark scores, assuming that the highest-scoring checkpoint will remain the best starting point for subsequent training. We show that this assumption can fail in a full 30B mixture-of-experts training pipeline. The checkpoints that perform better after the full downstream training stack also have higher solution density, i.e., retain downstream performance under local weight perturbations.
Read the full abstract ▾
All authors Sohir Maskey, Philipp Scholl, Jonas Knupp, Pit Neitemeier, Sascha Wirges
Artificial Intelligence
PlannerForge: LLM Agents for Scenario-Based Testing of Motion Planners in Autonomous Driving
Yuan Gao, Sebastian Müller, Mattia Piccinini et al. (+5 more)
Ensuring the safety of autonomous driving is a critical challenge. Scenario-based testing is a systematic process used to validate Autonomous Driving Systems (ADSs), but it remains a fragmented modular pipeline in which scenario generation,
retrieval, modification, ADS execution, and results analysis are performed by separate tools with little interaction. Large
Language Model (LLM)
agents have shown promise across ADS sub-systems such as perception, planning, and control. However, no prior work covers the whole scenario-based testing pipeline for ADSs with a unified LLM-
agent framework. We present PlannerForge, an LLM-
agent framework that extends all scenario-based testing stages (from Scenario Generation to ADS Assessment) and adds two further LLM-enhanced stages: ADS Enhancement and ADS Benchmarking. We evaluate PlannerForge with 10 off-the-shelf LLMs across all tasks (Generation, Selection, Modification, Module Routing, Planner Testing, and Enhancement) under 5
prompt conditions. Best-per-task scores range from 0.88 to 1.00, and open-source 20-35B backends match commercial APIs on most tasks. Open-source models such as Qwen3.6:35B match commercial APIs on three of the five tasks. Chaining the modules end-to-end retains 83% / 78% of seed queries (commercial / open). It outperforms Scenario Factory 2.0 (Finkeldei et al., 2025) on natural-
language generation (193 vs. 144 executable of 200) and realises 92-96% of requested city, road and vehicle attributes. It outperforms BM25 (Robertson and Zaragoza, 2009) at rank 1 selection (92.0% vs. 67.5%) and From-Words-to-Collisions (Gao et al., 2025) on physically valid edits (>=94% vs. 31%). At N=400, cost-tuning lifts planner success from 50.4% to 70.2% and cuts collisions from 19.0% to 8.4%, without domain-specific fine-tuning.
Read the full abstract ▾
All authors Yuan Gao, Sebastian Müller, Mattia Piccinini, Marc Kaufeld, Yuchen Zhang, Finn Rasmus Schäfer, Qunying Song, Johannes Betz
Artificial Intelligence · Cross-listed from cs.DB
SQLMorph: Query Mutation and Fine-Grained Metrics for Text-to-SQL Evaluation
Mohammadhossein Malekpour, Mohamed Riahi, Maxime Lamothe et al. (+1 more)
Text-to-SQL systems translate natural
language queries into executable SQL, democratizing access to structured data. Despite recent advances driven by large
language models (LLMs), evaluation remains a major bottleneck: public
benchmarks fail to capture the complexity of enterprise schema, while building private evaluation sets is costly and nondeterministic, making evaluation results difficult to reproduce. To address this issue, we present SQLMorph, a framework for Text-to-SQL evaluation via query mutation. SQLMorph introduces two techniques to automatically generate and expand evaluation sets: Join Query Expansion (JQE), which systematically increases structural complexity through valid join additions, and Textual Query Augmentation (TQA), which generates controlled natural
language perturbations to assess robustness to linguistic variation. JQE and TQA create targeted choke points to challenge specific system components. When applied to state-of-the-art systems, JQE increases query coverage and reveals accuracy degradation as the number of joins grows. Meanwhile, TQA shows that linguistic brittleness induced by heavy abbreviation can reduce accuracy by up to 17%. Beyond evaluation sets, SQLMorph introduces a family of execution-level metrics that address the limitations of current binary measures, such as Execution Accuracy. We define Execution Precision (EXP) and Execution Recall (EXR) to quantify the fraction of correct and recovered results, respectively, and combine them via F1 for unified scoring. Our experiments show that these relaxed metrics enable fine-grained analysis of over- and under-prediction, revealing differences across systems that binary metrics obscure. Together, SQLMorph's query mutation and fine-grained metrics support debugging and better align Text-to-SQL evaluation practices with real-world deployments.
Read the full abstract ▾
All authors Mohammadhossein Malekpour, Mohamed Riahi, Maxime Lamothe, Amine Mhedhbi
Artificial Intelligence
SkillAdam: Stable and Efficient Skill Evolution for Agents
Gaoyuan Li, Meihao Fan, Yizhe Liu et al. (+7 more)
Agent skills provide a lightweight way to equip frozen
language-model
agents with domain knowledge and procedural guidance, yet obtaining high-quality skills remains costly and difficult to scale. Expert-written skills require substantial human effort. Recent skill self-
evolution methods automate an iterative loop that uses execution feedback to revise skills, but their heuristic update strategies often yield unstable optimization and low iteration efficiency. We identify two challenges in realizing stable and efficient skill self-
evolution. Direction Stability requires effective corrections to accumulate rather than be overwritten by iteration-local feedback. Update Adaptivity requires the scope of each revision to reflect the consistency of recent case-level improvements. We introduce SkillAdam, an Adam-inspired framework for optimizing discrete and non-differentiable skill documents. As a functional analogue of Adam's first moment, an optimization
memory records identified problems and the outcomes of prior solution attempts to stabilize the update direction. As a functional analogue of Adam's second moment, a
volatility-driven edit budget tracks the history-weighted variation of recent case-level improvements and adaptively controls the update magnitude. Across seven
benchmarks that span short- and long-horizon tasks, SkillAdam achieves state-of-the-art performance with more stable optimization dynamics. It also obtains stronger skills with substantially fewer optimization iterations and lower cost than prior methods. Code repository: https://github.com/ruc-datalab/SkillAdam
Read the full abstract ▾
All authors Gaoyuan Li, Meihao Fan, Yizhe Liu, Shaolei Zhang, Ju Fan, Siyi Wang, Jiaheng Hou, Xudong Weng, Honghan Tian, Zang Li
Artificial Intelligence · Cross-listed from cs.CL
Evaluating and Improving Evidence-Grounded Fact-Checking in LLMs via Multi-Round Evidence Ablation
Xingyu Deng, Mingzi Cao, Nikolaos Aletras et al. (+2 more)
Automatic fact-checking systems assess the veracity of claims given evidence from relevant documents. Large
Language Models (LLMs) have demonstrated strong performance in fact-checking due to their general
reasoning capabilities. However, it remains unclear whether they faithfully make use of the evidence provided to reach veracity judgments or rely on parametric knowledge. To investigate this, we introduce Fact-Ablated Evaluation (FAE), a new evaluation framework that iteratively ablates the cited evidence to assess whether LLMs revise their predictions accordingly. Our empirical results show that current off-the-shelf LLMs as fact-checking systems rely more on their parametric knowledge than on the evidence provided. To bridge this gap between prediction accuracy and evidence grounding, we propose REAL (Rigorous Evidence Ablation Learning), a training framework that promotes evidence-dependent verification through counterfactual evidence supervision for the LLM-as-verifier models. Experiments on four fact-checking
datasets across different domains demonstrate that models trained with REAL obtain superior evidence-dependent capabilities compared to standard fine-tuned models. Our findings highlight that strong fact-checking performance can still coexist with weak evidence dependency, while REAL encourages veracity predictions to remain more closely tied to the availability of supporting evidence.
Read the full abstract ▾
All authors Xingyu Deng, Mingzi Cao, Nikolaos Aletras, Xi Wang, Mark Stevenson
Artificial Intelligence · Cross-listed from cs.ET
OntoKG-EQ: A provenance-grounded, competency-question-governed knowledge graph for auditable analyst querying
Furqan Nasir, Muhammad Atif Saeed, Muhammad Ehsan et al. (+2 more)
Analysts in emerging equity
markets keep answering the same questions. Did fundamentals match the
market's response? How does the local currency co-move with returns? Which firms outperform sector and
benchmark, and which disclosures coincide with abnormal trading? These answers come from ad-hoc spreadsheets that are hard to reproduce, audit, or trust. We present OntoKG-EQ, a knowledge-based system that makes such queries reproducible, evidence-linked, temporally explicit, valid, and inspectable. It couples a bounded, competency-question-governed core ontology with a provenance-aware knowledge graph in which every class, property, shape, and metric is justified by one of five frozen questions. The system materialises
market data into the graph, computes the metrics, validates its structure against declarative shape constraints, answers each competency question with a graph query, derives typed findings, and generates an explanation tracing each result to its observations, evidence, sources, and provenance. We evaluate on curated
datasets from three emerging
markets (Pakistan, Malaysia, Indonesia). Once each
market's data is mapped into the common schema, the ontology, shapes, queries, and rules are reused unchanged. A relational-database baseline shows the graph changes no analytics. Its value is governance, provenance, and self-explaining structure. Because answers are rendered deterministically from the validated graph, their consistency with it is guaranteed by construction. Used as a reference, the system measures how consistently eight open
language models transcribe the same evidence (provenance coverage 0.00 to 1.00). A study with a 17-participant convenience panel finds the evidence bundle significantly increased perceived trust and completeness. Code and data are openly released.
Read the full abstract ▾
All authors Furqan Nasir, Muhammad Atif Saeed, Muhammad Ehsan, Sher Jeel Ahmad, Abdul Moiz Altaf
Artificial Intelligence
API Benchmark Scores Do Not Reliably Transfer to Chatbot Interfaces
Jennifer Wang, Joachim Baumann, Daniel E. Ho et al. (+1 more)
Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trust, and influence policy. Yet, a key assumption underlying
benchmark scores is that the model performance measured through APIs faithfully reflects the behavior of deployed systems. We challenge this assumption by
auditing ChatGPT, Claude, and Gemini across seven systems and nine
benchmarks spanning general capability, social bias, and sycophancy. We find systematic API–interface differences in both accuracy and consistency. On average, API evaluations score 3.4 percentage points higher in accuracy and 2.1 percentage points higher in test–retest agreement than corresponding interface evaluations. For ChatGPT, the performance difference between API and interface access rivals the API-only difference between GPT 5.3 and GPT 5.4. Put differently, switching access surfaces can degrade performance as much as downgrading a full model version. We further test whether exposed API controls can reproduce interface behavior by varying system
prompts, sampling parameters, and
reasoning settings. These controls shift behavior in some cases but do not reliably eliminate the gap. Our findings document a context-validity gap: measurements obtained through APIs do not necessarily generalize to corresponding deployed interfaces, complicating the use of API evaluations as proxies for deployed systems.
Read the full abstract ▾
All authors Jennifer Wang, Joachim Baumann, Daniel E. Ho, Sanmi Koyejo
Artificial Intelligence · Cross-listed from cs.LG
Earth System World Model for What-If Simulations: A Case Study for Terrestrial Ecosystems
Zhihao Wang, Ruichen Wang, Ruohan Li et al. (+6 more)
Machine learning emulators have become essential for accelerating expensive Earth-system simulations, but most existing approaches remain passive forecasters: they reproduce simulator trajectories under prescribed forcings without an explicit interaction mechanism for user-specified interventions. This limits their use in interactive scientific workflows and Earth-system digital twins, where users often need to explore how a system would respond if selected state components were changed. We propose an action-conditioned world-modeling framework for Earth-system emulation that reformulates simulator trajectories as supervision for controllable state-transition learning. The key idea is transition-action pretraining: naturally observed state changes are treated as label-free action supervision, allowing the model to learn both prescribed dynamics and action-conditioned responses without manually annotated interventions. We further introduce masked response learning to infer unobserved variables under partial state edits and learn coupled system dependencies. We test this framework on ecosystem dynamics across six global regions and multiple stand ages. Experiments show that the model preserves competitive long-horizon emulation accuracy while enabling controllable structural interventions and coherent responses in coupled ecosystem-cycle variables. These results suggest a practical route from passive Earth-system emulators toward interactive, intervention-aware scientific surrogates.
Read the full abstract ▾
All authors Zhihao Wang, Ruichen Wang, Ruohan Li, Lei Ma, George Hurtt, Xiaowei Jia, Gengchen Mai, Shaowen Wang, Yiqun Xie
Artificial Intelligence
Closing the Consistency Gap: Self-Evolving Agents That Learn to Stay on Course
Evelyn Duesterwald, Benjamin Elder, Lilian Ngweta et al. (+2 more)
Large
language model (LLM)-powered
agents can be accurate on average yet unreliable in production, a discrepancy that has been observed but remains largely unaddressed. When given the same task five times, a ReAct
agent on the AppWorld
benchmark using GPT-4.1 succeeds in all five runs only 53% of the time, even though its per-run pass rate averages 77%. We call this 24-point shortfall the consistency gap, and we argue that addressing it is a precondition for trustworthy AI
agent deployment. We present a self-evolving
agent framework that reduces this gap by identifying unstable, low-consistency steps in
agent trajectories and converting them into episodic
memory the
agent can draw on in future runs. At its core is a Consistency Analyzer that pinpoints where and why a trajectory is likely to flip across executions, and a Guideline Generator that converts the diagnosis into targeted guidelines, committed to
memory and injected into future
agent executions on similar tasks. On AppWorld with ReAct/GPT-4.1, our framework raises the fraction of tasks that succeed in all five runs by +16 points on same-task evaluation and +13 points on similar-task generalization.
Read the full abstract ▾
All authors Evelyn Duesterwald, Benjamin Elder, Lilian Ngweta, Shashanka Ubaru, Malgorzata Zimon