DCGC is a Masked Diffusion Model framework that performs global correction of flawed reasoning traces in Large Language Models. It uses an imperfect solution draft from an upstream solver as auxiliary context and combines task‑specific supervised fine‑tuning with a Dynamic Dual‑CFG inference mechanism that separates problem‑only and joint problem‑draft branches. Experiments on math, code, and knowledge reasoning benchmarks show that DCGC outperforms standard sampling and simpler CFG variants, and it can improve full test‑set accuracy even when ground‑truth failure labels are unavailable.
By Minhae Oh, Nakyung Lee, Jungwoo Lee
The paper introduces a method to reduce sycophancy in large language models by using the Bayesian Truth Serum (BTS) as a reward signal in Group Relative Policy Optimization (GRPO). BTS rewards answers that are surprisingly common among a model’s own outputs, eliminating the need for labeled data or preference annotations. Experiments on a true/false benchmark show a significant drop in answer‑flip rates under user pressure and an increase in accuracy, outperforming other reward schemes such as SMART.
By Serhii Mytsyk, Yiming Zhang, Vikram Krishnamurthy
TAU-Agent is a retrieval‑augmented framework designed for traffic anomaly understanding in transportation videos. It uses a central retrieval agent to coordinate a Video Captioning Tool and an Open‑Vocabulary Tracking Tool, gathering captions, temporal intervals, and object trajectories relevant to a query. The collected evidence, along with sampled frames and the query, is fed into a fine‑tuned vision‑language model that reasons and generates an answer. TAU-Agent was evaluated on the AI City Challenge 2026 benchmarks, achieving notable scores across multiple tracks and ranking second, twelfth, and fifth respectively.
By Yuqiang Lin, Yan Shi, Sam Lockyer, Harish Tayyar Madabushi, Adrian Evans, Wenbin Li, Yinhai Wang, Nic Zhang
The paper introduces MAMA-FLUX.2, a conditional latent flow‑matching model built on FLUX.2-Klein-4B, designed to synthesize post‑contrast breast DCE‑MRI from pre‑contrast images for the MAMA‑SYNTH challenge. It encodes the pre‑contrast image as spatial conditioning and predicts the flow field for the post‑contrast latent, adapting the pretrained model with LoRA fine‑tuning and a regional training objective that combines global flow matching, tumor‑region supervision, and stable foreground regularization. Ablation studies show that moderate tumor and stable‑foreground weighting improves the balance between image fidelity and tumor‑region accuracy, with the best configuration achieving a strong trade‑off using LoRA rank 64/64, MHA_max 25, λ_tumor 0.25, and λ_stable 0.1.
By Kamil Kwarciak, Marek Wodzinski
The paper introduces Efficient Retrieval Adapter (ERA), a query‑side adapter framework that enables dense retrieval systems to adapt to asymmetric query–document scenarios without re‑indexing. ERA first aligns the embedding spaces of a powerful query embedder and a lightweight document embedder using unlabeled documents, then fine‑tunes the aligned query representation with a small set of labeled query‑document pairs. In experiments on 126 MAIR retrieval tasks across six domains, ERA boosts average nDCG@10 by up to 8.2 points in symmetric settings and over 12 points in asymmetric settings while requiring far fewer labels than fully supervised adapter training.
By Seiji Maekawa, Moin Aminnaseri, Pouya Pezeshkpour, Estevam Hruschka
MoganBert-TR is a 149‑million‑parameter Turkish encoder foundation model trained from scratch on a language‑specific corpus using a two‑stage CLM‑to‑MLM curriculum. The model, along with its embedding variant MoganBert‑Embed, achieves state‑of‑the‑art results on Turkish benchmarks such as TrGLUE and TabiBench, outperforming existing Turkish BERT models. Its tokenizer, comprising 50,048 tokens, also surpasses other Turkish tokenizers in compression and fertility metrics.
By Furkan Yilmaz, Habibe Aleyna Tasdemir, Muhammed Faruk Gozay
StoSignSGD is a new sign‑based optimization algorithm that injects structural stochasticity into the sign operator, ensuring unbiased updates. It resolves the divergence issues of traditional SignSGD on non‑smooth objectives, achieving optimal convergence rates in convex settings and improved complexity bounds in non‑convex, non‑smooth problems. Empirical results show that StoSignSGD is stable and efficient across large language model training, outperforming AdamW and SignSGD in low‑precision regimes (FP8 and FP4) and delivering speedups and accuracy gains on models ranging from OLMo2‑370M to 7B LLMs.
By Dingzhi Yu, Rui Pan, Yuxing Liu, Difan Zou, Tong Zhang
The paper introduces SAME (Stabilized Mixture-of-Experts) to address challenges in Multimodal Continual Instruction Tuning (MCIT) for large language models. SAME mitigates router drift by decomposing routing dynamics into orthogonal subspaces and updating only task-relevant directions, while preventing expert drift through curvature‑aware scaling that uses historical input covariance without rehearsal. The method also employs adaptive expert activation to freeze selected experts during training, reducing redundant computation and cross‑task interference, and demonstrates state‑of‑the‑art performance on a new long‑task‑sequence benchmark.
By Zhen-Hao Xie, Jun-Tao Tang, Yu-Cheng Shi, Han-Jia Ye, De-Chuan Zhan, Da-Wei Zhou
The paper introduces AlgoTrace, a framework that traces and steers algorithmic operations in large language models’ latent space during multi‑step reasoning. By clustering latent activations on tasks such as TSP, 3SAT, AIME, and Graph Navigation, the authors identify reusable primitive vectors that can be injected to elicit specific algorithmic behaviors, composed algebraically, and transferred across models and tasks. Fine‑tuning further improves the composition of these primitives, suggesting that LLM reasoning can be viewed as a walk through algorithmic primitives governed by compositional geometry.
By Samuel Lippl, Thomas McGee, Kimberly Lopez, Ziwen Pan, Pierce Zhang, Salma Ziadi, Oliver Eberle, Ida Momennejad
MeMark introduces a watermarking scheme for Spiking Neural Networks that embeds a multi‑bit identifier directly into the membrane state of selected Leaky Integrate‑and‑Fire neurons, rather than in the output head. The watermark is recoverable by comparing neuron firing thresholds, eliminating the need for a learned decoder. Experiments on various SNN architectures—including a 215.4M‑parameter SpikeGPT checkpoint—show that all 20 independent 64‑bit keys reliably pass verification under a 51/64 rule, remain robust after fine‑tuning, pruning, quantization, and output‑head replacement, and are not recovered by random keys or adaptive attacks within the tested threat model.
By Roberto Ria\~no, Gorka Abad, Stjepan Picek, Aitor Urbieta
The paper introduces FairTPT, a fairness-aware test‑time prompt tuning method for vision‑language models like CLIP. It jointly minimizes target marginal entropy while maximizing spurious marginal entropy to reduce bias under subpopulation shifts. Experiments show that standard episodic test‑time adaptation can worsen disparities, but FairTPT outperforms existing debiasing methods while preserving overall performance.
By Yoann Launay, Parameswaran Kamalaruban, Tom Kempton, Stuart Burrell, David Sutton
arXiv:2607.18227v2 Announce Type: replace
Abstract: In line with the prevailing direction of vision research, we explore the integration of both generation and editing capabilities for video and imag...
By Dingyun Zhang, Lixue Gong, Wei Liu
The paper investigates how refusal training shapes the internal geometry of language models, showing that activation updates from refusal-completion losses create a distinct low‑dimensional refusal subspace. In a case study on OLMo‑2‑0425‑1B‑Instruct, the authors link the brittleness of refusal directions to repetitive refusal prefixes and demonstrate that using diverse refusal starts can increase the stable rank of gradients, thereby hardening the model against vector‑ablation attacks. The work provides insights into the emergence of safety‑critical features and offers a potential strategy to strengthen refusal robustness.
By Andrey Labunets
arXiv:2608. 25826v1 Announce Type: cross Abstract: A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web passages, recovers only local thoughts, and leaves the structure of whole documents untouched.
By Qiankai Xu, Qiguang Chen, Zixin Su, Wenhao Huang, Yue Gao, Jiaheng Liu, Ge Zhang
The paper introduces V‑Rubrics, a reinforcement‑learning framework that evaluates vision‑language model responses by breaking them into atomic propositions and scoring them on Visual Faithfulness, Reasoning Consistency, and Instruction Following. Using a fine‑tuned Qwen3‑VL‑8B‑Instruct model and a newly created 50K‑example V‑Rubrics dataset, the authors demonstrate that rubric‑based GRPO outperforms both a shared SFT baseline and an answer‑only GRPO, especially on knowledge‑oriented and visually grounded reasoning tasks.
By Shulin Tian, Minglun Li, Yuhao Dong, Hao Ding, Jiarui Yao, Haiwen Diao, Jingkang Yang, Hongyuan Zhu, Ziwei Liu
The paper introduces Semantic Boundary Predictor (SBP), an inference‑time framework that improves demographic fairness in synthetic face generation by applying a single, one‑shot intervention during reverse denoising. SBP learns linear semantic boundaries from late‑stage latent representations and applies them only at the initial noisy latent, leaving the rest of the diffusion process unchanged. Experiments on CelebA‑HQ show significant reductions in fairness disparity—98% for gender, 95% for binary race, and 15% for four‑class race—while preserving image quality across demographic groups.
By Subir Kumar Parida, Rajbabu Velmurugan, Ketan Kotwal, R. S. Sengar, Swati Hiremath
PreResQ‑R1 introduces a Preference‑Response Disentangled Reinforcement Learning framework for Visual Quality Assessment that jointly optimizes absolute score regression and relative ranking consistency. It employs a dual‑branch reward system—modeling intra‑sample response coherence and inter‑sample preference alignment—trained with Group Relative Policy Optimization. The method extends to video quality assessment via a global‑temporal and local‑spatial data flow strategy, achieving state‑of‑the‑art results on 10 IQA and 5 VQA benchmarks with only 6K images and 28K videos, and provides human‑aligned reasoning traces.
By Zehui Feng, Weichuan Wang, Xiaohan Chen, Ting Han
OVO‑S‑Bench is a fully human‑annotated benchmark designed to evaluate streaming spatial intelligence in multimodal large language models (MLLMs). It contains 1,680 questions derived from 348 source videos, each with a query timestamp and evidence interval, and tests models on four levels of abstraction: instantaneous egocentric perception, spatiotemporal context tracking, generative spatial reasoning, and allocentric spatial mapping. Across 38 MLLMs, Gemini‑3.1‑Pro scored 59.2 versus 92.2 for human experts, with allocentric spatial mapping identified as the main challenge, and the benchmark reveals that chain‑of‑thought reasoning can worsen spatial errors when not grounded in the stream.
By Yifei Li, Pengyiang Liu, Yuhang Zang, Zhongyue Shi, Qi Fu, Hongye Hao, Jiwen Lu
The paper introduces ReT-2, a unified retrieval model that handles multimodal queries containing both images and text and searches across multimodal document collections. It employs a recurrent Transformer architecture with LSTM-inspired gating to integrate information across layers and modalities, capturing fine-grained visual and textual details. Evaluations on M2KR and M-BEIR benchmarks show state‑of‑the‑art performance, faster inference, and lower memory usage, and the model also boosts downstream tasks in retrieval‑augmented generation pipelines.
By Davide Caffagni, Sara Sarto, Marcella Cornia, Lorenzo Baraldi, Rita Cucchiara
The paper presents a training‑free, agentic computer‑vision system that converts 2D structural framing plan PDFs into editable finite‑element model drafts. It uses a deterministic stage to extract geometric primitives, estimate scale, and recognize five entity classes via a drafting grammar, followed by an agentic stage that applies typed corrections and fail‑closed transactions. Evaluation on a 100‑plan benchmark shows high precision and recall across columns, beams, walls, braces, and openings, with scale estimates within 0.1% of reference.
By Mohammad Talebi-Kalaleh, Qipei Mei