Tracked topic
…This view motivates FeatCal, which uses a small calibration set to calibrate the merged model weights layer by layer in forward order , reducing feature drift while staying close to merged weights and…
…However, little is known about the frequently used chat template 's effect on the calibration of conversational LLMs. In this work, we investigate the mechanisms driving this mis calibration by decoupling the…
…XiuYu Zhang , , , Abstract Self-Evaluation Elicitation (SEE) method improves model calibration for quality assessment through calibration-coupled reinforcement learning and masked distillation, demonstrating transferable quality evaluation beyond specific judge preferences. Generated by…
…To give agents a more complete picture of the state space they are operating in, we introduce calibrated verifier telemetry (CalVerT), which augments the agent's state with additional telemetry: a calibrated…
Papers arxiv:2606.16429 Taylor-Calibrate: Principled Initialization for Hybrid Linear Attention Distillation Published on Jun 15 Submitted by ZHONGZHU ZHOU on Jun 19 Together Authors: , , , , , , Abstract Hybrid linear attention models can…
Papers arxiv:2605.00323 Online Self-Calibration Against Hallucination in Vision-Language Models Published on May 1 Submitted by steven young on May 4 Authors: , , , , , Abstract Online self-calibration framework using Monte…
…Typed Hallucination Auditing and Calibrated Multi-Agent Debate for Trustworthy Legal AI Published on Jun 16 Submitted by Lalit Yadav on Jun 19 Independent Research Authors: Lalit Yadav , Akshaj Gurugubelli Abstract LegalHalluLens…
…A system may be built from CAD renderings, simulated RGB-D observations , normal reference images, synthetic defects , pretrained feature spaces , or language prompts, yet deployed under different sensors, lighting, materials, fixtures, calibration…
…For generators released after training, we introduce anchor calibration , a post-hoc, per-system Bradley-Terry calibration that recovers agreement at substantially better data efficiency than from-scratch retraining. The same frozen…
…To rigorously diagnose grounding behavior, we propose Calibrated Accuracy (mathrm{Cal)}, which jointly rewards correct grounding and appropriate abstention, incorporating absent-persona queries within a unified evaluation framework. On our dedicated experiments…