arXiv subjects

Dong Le

Publications and source records attributed to Dong Le.

3 recordsLinked to original sources

arXiv2026

GRACE: Step-Level Benchmark for Faithful Reasoning over Context

Many reasoning tasks require models to reason over input context, from document-grounded question answering to rule-based deduction. Chain-of-Thought (CoT) prompting produces traces that appear transparent, yet individual steps can silently deviate from the source evidence, even when the final answer is correct. Existing methods detect hallucinations at the response level but fail to identify where in the chain a failure occurs or what type it is. We introduce GRACE, the first human-annotated step-level faithfulness benchmark with a data-driven error taxonomy for context-grounded textual reasoning. GRACE covers CoT traces from 10 models across 4 source datasets, with each step annotated for faithfulness, error category, and natural language explanation. A data-driven taxonomy, discovered bottom-up via unsupervised clustering, organizes failures into two tracks: GRACE-Inference (deductive errors) and GRACE-Grounding (factual grounding errors), with four categories each. The evaluation set is human-annotated and challenging by design. Our experiments reveal substantial headroom for current models. In addition, integrating step-level faithfulness signals into reinforcement learning pipelines improves both downstream accuracy and reasoning reliability.

cs.CL↗

arXiv2026

Don't Read Everything: A Curvature-Conditioned Query for Linear Attention

Linear attention reduces the quadratic cost of softmax attention by maintaining a recurrent fast-weight state, but it consistently lags on in-context retrieval and long-context tasks. Existing remedies act on the write side of memory through gating, delta updates, or kernel feature maps, but the read step is left unchanged: every past key contributes additively to the output, so useful targets are diluted by the bulk of stored vectors. We borrow one specific piece of softmax's geometry to construct a cheap read-time contraction of the query. A second-order Taylor expansion of the softmax log-partition at the isotropic-attention point gives a local quadratic model whose curvature coincides with the running key covariance, a quantity that can be maintained with the same recurrent/chunkwise mechanism as the linear-attention state. The associated linear operator contracts the query along the high-variance directions of memory before it reads the state. We call this mechanism Curvature-Conditioned Query (CCQ). CCQ modifies only the read step and is composable with any linear-attention backbone. Attached to GLA and Gated DeltaNet, it improves perplexity, zero-shot downstream accuracy, S-NIAH retrieval at and beyond the training context, length-extrapolation perplexity from 4K to 20K, and LongBench accuracy.

cs.CL↗

arXiv2025

Spectroscopic Determination of Site-Selective Ligand Binding on Single Anisotropic Nanocrystals

Organic surface ligands are integral components of nanocrystals and nanoparticles that have a strong influence on their physicochemical properties, their interaction with the environment, and their ability to self-assemble and order into higher-order structures. These hybrid nanomaterials are tunable with applications in catalysis, directed self-assembly, next-generation optoelectronics, and chemical and quantum sensing. Critically, future advances depend on our ability to rationally engineer their surface chemistry. However, fundamental knowledge of ligand-nanoparticle behavior is limited by uncertainty in where and how these ligands bind to surfaces. For nanoparticles, in particular, few characterization techniques offer both the high spatial resolution and the precise chemical mapping needed to identify specific ligand binding sites. In this study, we utilized synchrotron infrared nanospectroscopy (SINS), atomic force microscopy (AFM), and scanning tunneling microscopy (STM) together with first-principles computer simulations to validate the site-selective adsorption of organic ligands on a shaped nanocrystal surface. Specifically, we demonstrate that the sterically encumbered isocyanide ligands (CNAr^{Mes2}) preferentially bind to the high curvature features of Ag nanocubes (NCs), where low-coordinate Ag atoms are present. In contrast, isocyanide ligands that do not exhibit these steric properties show no surface selectivity. SINS serves as an effective tool to validate these surface binding interactions at the near-single molecule level. These results indicate that steric effects can be successfully harnessed to design bespoke organic ligands for fine-tuning nanocrystal surface chemistry and the properties of the nanocrystal ligand shell.

cond-mat.mtrl-sci↗