Relation
-
Where's the Indirect Evidence? Investigating Grammatical Generalization with Influence Functions
2026-10-16 04:32:36William Dai, Zhenghao Herbert Zhou, Maya Viswanathan, Robert FrankAbstract
Language models generalize beyond their training corpus to learn linguistic phenomena (Patil et al., 2024; Misra and Mahowald, 2024), showing that there exists indirect evidence driving learning. Prior work has inferred that indirect evidence exists without identifying its character or considered specific hypotheses about what it might be. We explore an alternative bottom-up method for identifying indirect evidence. We use influence functions (Koh and Liang, 2017) to measure how individual training sentences affect model predictions in 15 grammatical phenomena from BLiMP (Warstadt et al., 2020). We filter out a phenomenon's direct evidence using hand-constructed rules and take the 60,000 most influential remaining sentences as indirect evidence. This indirect evidence substantially affects learning: retraining without it lowers held-out accuracy more than removing a matched random group for every phenomenon across five models (by a median of 12 points). We characterize the indirect evidence by its similarity to our direct evidence filters: it (partially) meets the filters' structural conditions at a median of 4.7 times the rate of a matched random group. Finally, clustering phenomena by indirect-evidence overlap recovers three of the five BLiMP categories that include multiple phenomena, suggesting that this method can reveal linguistic relationships.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22919
PDF
-
When to Forget: Risk-Controlled Context Adaptation for Long-Horizon Agents
2026-10-16 04:01:00Chenle Chen, Yangbo Wei, Zhijie Zheng, Zhenyu Su, Qifan Wang, Junhong Qian, Chao Yao, Chen Wu, Lei HeAbstract
Long-horizon agents learn from failure, but the value of a failure record can change as execution progresses. We study when retaining, compressing, or removing the same history block improves subsequent execution relative to full-history KEEP. We introduce RECAP, a framework that combines matched context interventions, execution-relational prediction, and independent policy certification. Matched environment states and rollout seeds measure success and continuation prompt-cost differences; prefix-only relations between a failure block and the current execution propose promising edits. Independent task bundles then select a policy subject to harmful-intervention and non-positive-benefit risk constraints under a fixed measurement protocol. A five-seed longitudinal case holds both the failure block and its summary fixed: compression fails at an early state and becomes a cost-saving intervention at a later state. Across five backbones, RECAP produces positive paired-success saving estimates with backbone-specific reliability-efficiency tradeoffs. On Hunyuan-A13B, it edits 43.24% of retained checkpoints while matching KEEP's observed 91.89% continuation success and reducing unconditional mean continuation prompt cost by 4,619.8 tokens (2.13%). The framework provides an execution-level reference for deciding when to forget: retain history by its value to what comes next.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22897
PDF
-
Operon: Incremental Construction of Ragged Data via Named Dimensions
2026-10-16 02:57:09Sungbin Moon, Jiho Park, Suyoung Hwang, Donghyun Koh, Seunghyun Moon, Minhyeong LeeAbstract
Modern data processing workflows frequently encounter ragged data: collections with variable-length elements that arise naturally in domains like natural language processing, scientific measurements, and autonomous AI agents. Existing workflow engines lack native support for tracking the shapes and dependencies inherent to ragged data, forcing users to manage complex indexing and dependency bookkeeping manually. We present Operon, a Rust-based workflow engine that addresses these challenges through a novel formalism of named dimensions with explicit dependency relations. Operon provides a domain-specific language where users declare pipelines with dimension annotations that are statically verified for correctness, while the runtime system dynamically schedules tasks as data shapes are incrementally discovered during execution. We formalize the mathematical foundation for reasoning about partial shapes and prove that Operon's incremental construction algorithm guarantees deterministic and confluent execution in parallel settings. The system's explicit modeling of partially-known states enables robust persistence and recovery mechanisms, while its per-task multi-queue architecture achieves efficient parallelism across heterogeneous task types. Empirical evaluation demonstrates that Operon outperforms an existing workflow engine with 14.94x baseline overhead reduction while maintaining near-linear end-to-end output rates as workloads scale, making it particularly suitable for large-scale data generation pipelines in machine learning applications.
Abstract (translated)
URL
https://arxiv.org/abs/2511.16080
PDF
-
Measurement-Adaptive Light-Cone Transform for Robust Non-Line-of-Sight Imaging
2026-10-16 02:47:58Lianfang Wang, Xueying Liu, Junnan Dong, Qinghua Zhang, Jun Liu, Yong Wang, Yuping DuanAbstract
Practical non-line-of-sight (NLOS) measurements are affected by noise and deviations from idealized light transport, which can degrade reconstruction with conventional light-cone transform (LCT) methods. We propose the Measurement-Adaptive Light-Cone Transform (MALCT), a physics-guided framework that retains the efficient Fourier-domain inversion structure of LCT. {Within a unified variational formulation, MALCT combines the analytical response with a shared learned response through measurement-conditioned weighting and predicts a measurement-conditioned, frequency-dependent regularization map.} An attention-enhanced spatio-temporal feature extraction module captures long-range temporal dependencies and local spatial correlations to support reconstruction under low-signal conditions. Experiments on synthetic data and real measurements from different NLOS systems demonstrate improved reconstruction quality. Controlled evaluations further assess robustness to temporal-response broadening and reduced acquisition time. MALCT also preserves scene structures with spatial sampling grids down to $32\times32$, supporting its potential for NLOS imaging under limited acquisition budgets.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22847
PDF
-
VTrajEmbed: Aligning Text with Video Object Trajectories via Spatiotemporal Embeddings
2026-10-16 02:34:07Ruizhi Chen, Zining Fan, Yudi Guo, Jiang Liu, Zhiyu Lin, Zhongjiang He, Shouqian Shi, Hao SunAbstract
Fine-grained alignment between text and video object trajectories is a core challenge in multimodal understanding. Existing embedding models excel at global retrieval, yet their single global representation entangles multiple objects and their spatiotemporal dynamics, limiting direct object-level grounding. To address this issue, we propose VTrajEmbed, an embedding model that decomposes a video into object trajectories and aligns each trajectory embedding with text for object-level grounding through retrieval. Our main contributions: (1)Slow-Fast Trajectory Encoding, a dual-branch architecture that establishes trajectory representations by densely sampling low-resolution frames for high-frequency motion cues and sparsely selecting high-resolution frames for scene context, aggregated by a shared LLM backbone; (2)Mask-Guided Contrastive Learning, a training objective built upon these representations that couples an intra-group loss and a batch-level bidirectional loss under a relation mask, excluding unverified relations and preventing confirmed matches from becoming false negatives; (3)Text2Traj-Bench, a trajectory-level text-video benchmark with model-assisted construction and human-verified query--trajectory relations. Experimental results show that VTrajEmbed-5B improves text-to-trajectory and trajectory-to-text R@1 over Qwen3-VL-Embedding-8B by 15.48 and 18.05 percentage points, respectively, and achieves 49.48% average R@1 across five global retrieval benchmarks, compared with 42.86% for Qwen3-VL-Embedding-8B.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22842
PDF
-
ARBOR: Locating Reasoning Failures in Long Chains of Thought through Reasoning State Tracking
2026-10-15 22:39:54Huihan Li, Siyuan Wang, Naga Vamsi Ramana Dinavahi, Xiaohui Liu, Xiang RenAbstract
Large reasoning models solve complex problems through long chains of thought, in which prior work has documented failure modes such as overthinking, underthinking, and unsupported deductions. Existing methods detect these failures from local signals: textual similarity to nearby steps, switch tokens, or an LLM's judgment of a step's correctness. However, what a step contributes to the solution depends on the state that earlier steps have established, so identifying failures requires a global view that tracks that state across the trajectory. Therefore, we propose ARBOR, a reasoning-failure attribution framework that represents each trajectory as a tree of reasoning states, built as each step creates, verifies, prunes, or revisits a state, and attributes each failure to a relation between a step and the state preceding it. Using ARBOR, we refine the three failure modes into distinct subtypes that harm reasoning through different mechanisms and at different costs. We find that the tree state ARBOR maintains before a step separates failure steps from the at-risk steps at which the failure did not occur. We discuss how ARBOR integrates with existing mitigations at inference time and in training data. Code, data and interactive examples are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22717
PDF
-
Risk and Refusal: Mechanisms of LLM Responses to Suicidal Ideation
2026-10-15 22:34:45Drishti Goel, Soorya Ram Shimgekar, Jeongah Lee, Violeta J. Rodriguez, Daniel S. Brown, Ravi Karkar, Koustuv SahaAbstract
Refusal is a common safety mechanism in large language models (LLMs), yet in contexts of suicidal ideation (SI), refusing to engage can itself constitute a safety failure. We evaluate how eight instruction-tuned LLMs respond to SI and self-harm disclosures across three datasets, varying whether models are informed that a disclosure expresses SI and whether they receive a clinician-developed response protocol, and complement this with mechanistic analyses in six open-weight models. Our results reveal a nuanced relationship between refusal and response-level risk. At baseline, elevated risk in model responses is driven primarily by inattention to the disclosure, alongside refusal to engage. Mechanistic analyses suggest that models may refuse SI disclosures through the same direction that mediates their refusals of harmful requests. However, informing models of the SI context, or suppressing refusal directly, increases engagement without reliably lowering response risk, and may raise it. In contrast, the clinician-guided protocol consistently reduces response-level risk while also reducing refusal. Together, these analyses and ablations motivate the need for models that refuse only when warranted and engage in interactionally safe ways when users disclose SI.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22715
PDF
-
Brain-Aligned Multimodal LLMs for Social and Emotional Understanding of Cartoon Movies
2026-10-15 22:22:03Yiran Huang, Zeyong Zhang, Bharat Biswal, Xin Di, Mengjia XuAbstract
Multimodal large language models (MLLMs) offer scalable annotations of social and affective content, but their temporal agreement with human judgments and neural predictive value require further evaluation. We compared diffusion-based and autoregressive MLLMs with continuous human Theory of Mind (ToM) and affective valence ratings across three short cartoon movies. Behavioral analyses assessed temporally aligned Pearson correlations, rating distributions, and the effects of preceding scene descriptions. Ratings from each source were separately used in subject-specific regression models to predict regional fMRI responses. Model ratings captured aspects of human temporal variation despite systematic distributional differences, with generally stronger alignment for valence than ToM. Preceding context improved alignment in some conditions, depending on the model and movie. In \emph{The Present}, more regions showed significant encoding for MLLM ratings than for human ratings, whereas effects were more limited in the other movies. Spatial overlap across rating sources occurred mainly among MLLMs; overlap between human and model ratings was limited and involved primarily visual, dorsal-attention, and posterior default-mode regions. These findings support MLLMs as scalable tools for generating brain-predictive annotations of social and emotional content, while highlighting systematic rating differences and model- and stimulus-dependent alignment with human judgments.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22706
PDF
-
Task Structure and Scalability in Multi-Robot Authoring
Abstract
Multi-robot authoring can reduce completion time by distributing drawing work, but the resulting speedup depends on task structure and the allocation-routing pipeline. We investigate these relationships by representing drawings as geometric stroke graphs and comparing component-level (single-token) allocation with shared-token (decomposed) allocation, which permits multiple robots to work within a connected component. A simulation study of 385 drawings and teams of one to eight robots shows diminishing speedup gains under single-token allocation and larger mean speedups under the shared-token pipeline, each normalized by its own single-robot completion time. Shared-token execution also exhibits lower idle-time shares at larger team sizes, although it does not reduce completion time for every drawing. Comparisons across structural attribute groups reveal associations between structural descriptors and differences in scaling behavior, with greater separation between several attribute groups at larger team sizes. These findings characterize empirical scaling in the evaluated pipelines and highlight the importance of considering within-component workload and graph structure alongside component count. Additional materials are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22670
PDF
-
From Relevant Memories to Sufficient Evidence: Rethinking Memory Utilization in LLM Agents
2026-10-15 21:34:17Shixuan Li, Wei Yang, Peiyu Zhang, Anzhe Cheng, Heng Ping, Xiaole Zhang, Jesse Thomason, Xuezhe Ma, Yue Zhao, Paul BogdanAbstract
Long-term memory enables LLM agents to reuse past experiences, yet retrieving relevant memories does not necessarily provide sufficient evidence for answering a query. Essential information may remain fragmented across memory records, obscured by abstraction, or disconnected from the relations required for reasoning. We identify this relevance-sufficiency gap and propose QEMem, a framework that reformulates memory utilization as the construction of query-specific evidence sets, shifting the focus from individual retrieval relevance to collective evidential support. QEMem integrates three complementary mechanisms: query-guided discovery retrieves evidence across multiple memory granularities; source-grounded reification organizes it into structured ViewCards with explicit evidential roles and source references; and sufficiency-guided refinement selectively recovers missing information. Operating on existing memory stores without modifying memory formation, QEMem achieves 91.43% LLM-judge accuracy on LoCoMo with Claude Opus 4.5, outperforming the strongest evaluated baseline, MemR$^3$, by 7.6 percentage points and NEMORI by 10.7 points under a shared memory store. Further analyses demonstrate improved evidence coverage and strong performance on questions requiring cross-session evidence integration.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22668
PDF
-
Inductive Learning of Concept Embeddings from Cross-Linguistic Colexification Patterns with Textual Attributes
Abstract
The advent of distributional semantics as a framework for computational semantics marked a paradigm shift in the nature of representations, by which meanings are represented as dense vectors commonly known as word embeddings. Computational approaches to historical or typological language comparison, on the other hand, still mostly rely on discrete representations, while the potential of embedding-based methods remains largely untapped in this context. Recent studies have proposed concept embeddings as a language-independent, onomasiological counterpart to word embeddings, based on networks of cross-linguistic colexification patterns. While the resulting concept embeddings reflect meaningful semantic properties and relations, they suffer from a lack of scalability, since they rely on transductive embedding approaches which only learn embeddings for nodes present in the training data. In this study, we present an inductive framework for learning concept embeddings, combining the structural properties of full and partial colexification networks with semantic information from short concept descriptions. This framework allows the embedding of unseen concepts based on a short definition. We show that the resulting concept embeddings correlate with various cross-linguistic measures of similarity or relatedness, while generalizing well to new concepts that were absent from the training data.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22638
PDF
-
DINO Eats Geometry: Self-Supervised Sign Representations that Absorb Articulated Hand Geometry
2026-10-15 20:29:41Junyi Hu, Zhewen He, Haomian Huang, Zhenhua Li, Zhifei Li, Shuaihang Yuan, Yi FangAbstract
Handshape and orientation are geometric components of sign language, yet many sign encoders learn from appearance alone. We introduce SignDEG (DINO Eats Geometry), a temporal sign encoder that uses articulated hand geometry as a pre-training target and requires only video at inference. Frozen DINOv3 features of tracked hands and faces feed per-stream temporal Transformers. Geometric latent prediction (GLP) supervises an intermediate block with WiLoR latents, MANO articulation, and 3D joints, including at masked frames. Geometric relation transfer (GRT) matches cross-video frame-similarity distributions at the final block. Both objectives are learned jointly with temporal self-distillation; WiLoR and the geometry heads are then discarded. Under the SHuBERT public-data evaluation protocol, SignDEG reaches 17.9/24.5 BLEU-4 on How2Sign/OpenASL, compared with 16.2/23.2 for SHuBERT. Rank-1 adaptation improves ASL Citizen and Sem-Lex recognition by 0.02 Recall@1 each, and phonological classification improves across 16 attributes. Removing GRT while retaining GLP reduces How2Sign BLEU-4 by 1.4 and ASL Citizen Recall@1 by 0.034. The results support transferring articulated structure into an appearance encoder without running a hand reconstruction model at deployment.
Abstract (translated)
URL
https://arxiv.org/abs/2609.06296
PDF
-
Alignment of visual brain encoding models with individual perceptual similarity judgments
2026-10-15 19:13:07Daniel Chong, Haomiao Chen, Keith W Jamison, Raja Marjieh, Ilia Sucholutsky, Nori Jacoby, Amy KuceyeskiAbstract
Encoding models that predict brain activation responses to images have traditionally been trained to maximize accuracy when predicting observed brain data. However, their internal representations may not reflect individual human perception. In this work, we investigate whether we can fine-tune brain encoding models to better align their representations with individuals' similarity judgments. Using similarity judgments of 100 images from 8 participants in the Natural Scenes Dataset (NSD), we adapt a generic human similarity judgment model (DreamSim) separately for each participant to obtain personalized perceptual similarity models. These models are then used to obtain subject-specific similarity structure across images in their encoding model training sets. We fine-tune each participant's baseline encoder in two conditions, aligning its representations with either generic or personalized similarity judgments, with the same initialization and objective. We find that baseline brain encoders reflect within-subject perceptual similarity more strongly than measured fMRI responses and preserve intersubject correlations present in measured fMRI, but fail to reflect those of human perceptual similarity judgments. Fine-tuning with generic, population-level perceptual similarity information results in better alignment with human judgments than baseline encoders, but these models still fail to capture intersubject correlations of human judgments. However, fine-tuning with personalized perceptual similarity information results in models that are the most aligned with human perceptual similarity judgments and also reflect intersubject correlations of human judgments. Our framework provides an approach for bootstrapping human similarity judgment data into more individualized, human-aligned brain encoding models of vision.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22569
PDF
-
Who Do Perspectivist Models Predict Well? Annotator Agreement, Sparsity, and Performance
2026-10-15 19:12:48London Lowmanstone IV, Vaagishwaran Sivakumar, Anwesha Samaddar, Akul Mundada, Harnoor Kaur, Maria GiniAbstract
Successful completion of perspectivist natural language processing (NLP) annotation tasks requires algorithms to be aligned across diverse perspectives in order to predict how different humans will rate text. This work studies how the performance of such algorithms varies for annotators who behave differently than others, either by providing annotations that diverge from others' responses or by providing fewer annotations altogether. To aid with characterizing model performance, we curate the largest collection to date of perspectivist NLP datasets with ordinal annotations. After evaluating multiple classes of perspectivist algorithms across each of the datasets, we model the relationship between individual annotator idiosyncrasy and algorithm performance. In general, we find that algorithms perform better on annotators who tend to agree with others, but that annotator response rate does not have a statistically significant positive correlation with algorithm performance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22567
PDF
-
Micro-Localization: High-Precision Mobile Robot Localization Using Microscopic Ground Texture
2026-10-15 19:04:52Aaron Wilhelm, Marina de Albuquerque Marcon, Emith Uyanwatte, Nils NappAbstract
We demonstrate high-precision global localization using natural ground texture viewed through microscope optics, achieving localization in the tens-of-micrometers regime. For the planar ground texture localization problem considered here, we relate optical magnification and image-space localization uncertainty to physical localization precision. Leveraging this relationship, we then develop a microscope-based localization system that addresses the calibration, camera-modeling, and focus challenges introduced by high-magnification imaging. We evaluate this system across several surface textures using a calibrated precision gantry. Across 1,007 held-out queries per surface, the system localizes every query and achieves translation RMSEs of 28$-$73 $\mu$m. Independent photometric registration agrees with the feature-based localization estimates to within approximately 9 $\mu$m RMSE. We further demonstrate localization on a low-cost mobile robot using the gantry-collected maps, and also show that the system can resolve micrometer-scale displacement caused by thermal expansion. These results show that microscopic natural surface texture can bring mobile localization into a precision regime typically associated with fixed positioning systems.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22563
PDF
-
HeRE: A Novel Benchmark for Hebrew Relation Extraction via LLM-Guided Denoising of Knowledge Graph Alignments
2026-10-15 18:40:22Ron Keinan, Amir David Nissan Cohen, Reut TsarfatyAbstract
While LLMs demonstrate strong capabilities in English Relation Extraction (RE) tasks, a significant gap remains in assessing their RE performance in low-resource languages, due to a lack of standardized benchmarks. Creating native, language-specific annotated RE benchmarks is expensive and requires human expertise, while relying on translation yields suboptimal, inaccurate datasets that introduce English-centric biases. To tackle this, we propose a scalable, language-agnostic methodology to create RE data by denoising naturally occurring yet noisy alignments of knowledge bases and corresponding texts. We define a proper alignment via a Natural Language Inference (NLI) task, where a relation triplet is entailed by an input text. We empirically assess five denoising strategies, and demonstrate that multilingual models outperform language-specific models in the few-shot LLM setting, while Hebrew-specific encoders remain competitive under NLI fine-tuning and are strongest in the label-free cross-training setting. We further find that prompting in the native language achieves performance on par with or slightly exceeding English prompts, confirming that the pipeline operates naturally in the target language. The best open-weights model, Gemma-4-31B-it (F1\,=\,0.915), matches the best API model, GPT-5.4 (0.921), and reaches F1\,=\,0.844 on a fully held-out, dual-annotated test set. Applying the optimized pipeline, we release HeRE: a novel, massive, 1.5-million-instance silver dataset for Hebrew RE, designed to support data-intensive experimentation and model fine-tuning. HeRE bridges a critical gap in standard Hebrew RE evaluation, while providing the NLP community with a reproducible, language-agnostic pipeline for RE dataset construction in other low-resource settings.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22542
PDF
-
DK-VLM: Geometry as Position for Spatial Reasoning in Vision-Language Models
2026-10-15 17:58:15Han Su, Meng Tian, Tianyu Huang, Mingyu Gao, Lennox Wei, Dongxu Li, Jianhua Han, Hang Xu, Xiaohe Wu, Wangmeng ZuoAbstract
Vision-language models still struggle to understand 3D spatial relations across views and update spatial judgments as the observer moves. A common approach fuses 3D reconstruction features into visual tokens, but aligning these features with pretrained semantic representations remains challenging under limited spatial supervision. We propose DK-VLM with Dual-Kernel Attention, which computes attention from both a visual token's input position and its geometric position in the scene. A semantic kernel retains the pretrained model's native attention and positional encoding, while a separate geometric kernel uses estimated 3D points and camera poses to directly modulate attention through point displacements and relative view transformations. Then, a learned gate fuses their outputs. Geometry thus enters attention as position rather than as fused reconstruction features. We further introduce Reference-Frame Composition (RC) supervision to train the model to update its reference frame across successive judgments. It provides stepwise turn and heading targets derived from verified single-step annotations. Among the compared methods, DK-VLM-8B achieves the highest average score across six spatial benchmarks and the highest overall scores on VSI-Bench (74.33) and ReVSI-32 (58.28). RC supervision further improves multi-step spatial reasoning, while the full model preserves general video understanding. Project page: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22356
PDF
-
AVC-Reward: Event-Centric Audio-Video Consistency Evaluation
2026-10-15 17:51:56Yuruo Wu, Kaihang Pan, Kai Shen, Jiacheng Li, Shuyuan Tu, Jisheng Dang, Li Tuoyu, Juncheng Li, Siliang TangAbstract
As joint audio-video generation models become more capable, they can generate audio-visual scenes exhibiting three forms of complexity: multi-event composition, multi-layer audio mixture, and asymmetric cross-modal relations. However, current evaluations largely reuse metrics from earlier simple and unidirectional tasks. They typically evaluate an audio-video clip as a whole and rely on simple predefined relations, making them unsuitable for these increasingly complex scenes. We propose an event-centric evaluation framework that disentangles audio and visual events at the event level, analyzes each event independently, cross-validates relations across modalities, and aggregates the judgments into a coarse-to-fine, three-level consistency metric. The metric assesses global narrative harmony, event-level temporal synchronization, and fine-grained attribute alignment, providing a general formulation for audio-video consistency evaluation across diverse generation settings and cross-modal relations. We instantiate the metric as AVC-Reward, a trainable audio-video reward model. To evaluate the agreement between automatic judgments and human perception across generation paradigms and relation complexities, we construct AVC-Benchmark, which is primarily built from diverse positive and negative samples generated by various types of audio-video generation models. AVC-Reward achieves state-of-the-art alignment with human judgment among automatic methods, provides interpretable rationales that localize failures, and improves the audio-video consistency of a joint generation model when used as an online reinforcement-learning signal. Project page: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.22321
PDF
-
Knowledge That Is Linear Before Fine-Tuning Generalizes After It
2026-10-15 17:47:52Faridoun Mehri, Hinrich Sch\"utzeAbstract
Not all facts a language model can recall support further learning equally well. We distinguish active facts, which support generalization beyond their own recall, from inert facts, which can be recalled but do not support such generalization in a given task. We study this distinction by constructing new relations from existing ones. Take the relation that maps a place to its continent (Paris -> Europe) and assign each of its answers an arbitrary new label (Europe -> crab, Asia -> zebra). The new relation then maps Paris -> crab. We fine-tune the model on the new relation for a subset of subjects, without telling it the original relation or the relabeling rule, and test whether it assigns the corresponding labels to held-out subjects that share an answer in the original relation. Having learned Paris -> crab, does it answer Munich -> crab? Across eleven language models and 70 relations, how linearly readable the original relation is in pretrained representations strongly predicts this generalization. Within each model, the two correlate across relations at Pearson r = 0.87 to 0.92. This association persists when controlling, separately and jointly, for the number of fine-tuning subjects and answers, the model's accuracy on the original relation, and how often the subjects occur in pretraining text. It is also robust to re-randomizing the label assignment and fine-tuning seed, and to replacing the animal labels with labels from thirteen other semantic categories. When we map a fine-tuned model's wrong label back to the original answer it was assigned to, Munich -> zebra becomes Munich -> Asia. Answers recovered this way are more often among the linear probe's top answers for the subject than among the pretrained model's own. Whether a new relation generalizes depends less on whether its facts can be recalled than on how readily their shared structure can be read out linearly.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22300
PDF
-
MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
2026-10-15 17:37:50Luozheng Qin, Yuhan Zhu, Changhai Zhou, Qian Qiao, Jun Gao, Yuhua Zhou, Harry Zhang, Xiangyu Zeng, Yuxuan Liu, Kaijie Chen, Andrew Chen, Pony MaAbstract
Spatial intelligence lets an agent perceive and act in the physical world, and many of its judgments must be made at interaction rates: which spatial relation holds, how far an object is, or where to place it. Vision-language models (VLMs) produce these answers by free-form text generation, paying for autoregressive decoding and answer parsing even though the output space is compact: one of a few options, a scalar, or a 2D point. In light of this, we introduce MJev-Spatial, a System 1 multimodal decision model that tackles all three types of spatial understanding tasks within a unified classification framework. Specifically, MJev-Spatial treats each candidate as a class and scores it using the backbone's pretrained query and key projections, without adding parameters. The candidates are the given options for choice tasks, or value bins and image regions refined through coarse-to-fine classification for numerical and point prediction, respectively. At inference, each classification decision takes a single forward pass and generates no answer tokens. To train on heterogeneous spatial data efficiently, we supervise all coarse-to-fine stages jointly with teacher forcing in one forward pass, and combine this with length-aware global batching and cross-GPU load balancing, raising training throughput by 30.8% over the original batching strategy. Built on Qwen3.5-4B and trained on 12.3M question-answer pairs converted from existing spatial datasets, MJev-Spatial achieves the highest overall score (65.64%) among eight tested models, ranks first on seven of twelve benchmarks, and outperforms its unmodified autoregressive base model on all twelve, while running 2.5 times faster than the fastest autoregressive baseline and 5.6 times faster than the base model.
Abstract (translated)
URL
https://arxiv.org/abs/2610.22276
PDF
