Evaluation-Conditioned Training: Teaching Models to Generalize to Stronger Oversight Regimes

• URL: http://arxiv.org/abs/2608.10209v1
• Date: Mon, 10 Aug 2026 20:22:13 GMT
• Title: Evaluation-Conditioned Training: Teaching Models to Generalize to Stronger Oversight Regimes
• Abstract summary: A key limitation of current post-training methods is the inability of human annotators and automated reward functions to faithfully capture the feedback we would like to give.<n>We introduce Evaluation-Conditioned Training (ECT), a post-training framework that uses natural language to condition each training sample on the fidelity of the feedback we provide.<n>ECT is aimed at improving performance under imperfect feedback and works as an add-on to existing algorithms such as SFT and PPO.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Feedback signals used to train Large Language Models (LLMs) are the primary driver of their behavior and our main lever for instilling alignment with human values and objectives. However, a key limitation of current post-training methods is the inability of human annotators and automated reward functions to faithfully capture the feedback we would like to give. We introduce Evaluation-Conditioned Training (ECT), a post-training framework that uses natural language to condition each training sample on the fidelity of the feedback we provide and then elicits the desired behavior by conditioning the LLM on a high-fidelity monitor in deployment. ECT is aimed at improving performance under imperfect feedback and works as an add-on to existing algorithms such as SFT and PPO. We first provide a conceptual framework for ECT and discuss its potential to address persistent sources of reward mis-specification. Then we motivate ECT in the context of the eliciting latent knowledge (ELK) problem. Finally, we evaluate ECT on two proof-of-concept experiments: increasing even-handedness in news article generation and reducing sycophancy on an arithmetic task. In each setting, we utilize imperfect feedback, rewarding bias and agreement with the user, respectively. In both settings, ECT improves the targeted behavior relative to direct training.

Related papers

• From Holistic Evaluation to Structured Criteria: Rubrics Across the Evolving LLM Landscape [79.30826980815927]
  rubrics are explicit criteria sets that transform complex quality judgments into structured and actionable standards.<n>We systematically organize existing rubric designs, examine their construction and optimization, and analyze their role across evaluation and training.
  arXiv  Detail & Related papers  (2026-06-07T13:34:55Z)
• RADAR: Revealing Asymmetric Development of Abilities in MLLM Pre-training [59.493415006017635]
  Pre-trained Multi-modal Large Language Models (MLLMs) provide a knowledge-rich foundation for post-training.<n>Current evaluation relies on testing after supervised fine-tuning, which introduces laborious additional training and autoregressive decoding costs.<n>We propose RADAR, an efficient ability-centric evaluation framework for Revealing Asymmetric Development of Abilities in MLLM pRe-training.
  arXiv  Detail & Related papers  (2026-02-13T12:56:31Z)
• Reshaping Action Error Distributions for Reliable Vision-Language-Action Models [69.38615670891038]
  In robotic manipulation, vision-language-action (VLA) models have emerged as a promising paradigm for learning generalizable and scalable robot policies.<n>We focus on continuous-action VLA models and move beyond conventional MSE-based regression by reshaping action error distributions during training.<n>We evaluate our approaches across standard, few-shot, and noisy settings on multiple representative VLA architectures.
  arXiv  Detail & Related papers  (2026-02-04T05:37:09Z)
• Reinforcement Learning from Implicit Neural Feedback for Human-Aligned Robot Control [0.0]
  We propose a novel reinforcement learning from implicit human feedback (RLIHF) framework that utilizes non-invasive electroencephalography (EEG) signals.<n>We evaluate our approach in a simulation environment built on the MuJoCo physics engine, using a Kinova Gen2 robotic arm.<n>Results show that agents trained with decoded EEG feedback achieve performance comparable to those trained with dense, manually designed rewards.
  arXiv  Detail & Related papers  (2025-11-18T22:50:32Z)
• FOSSIL: Harnessing Feedback on Suboptimal Samples for Data-Efficient Generalisation with Imitation Learning for Embodied Vision-and-Language Tasks [45.65159253753118]
  This work explores how agents trained with imitation learning can learn robust representations from both optimal and suboptimal demonstrations.<n>We provide language feedback embeddings as part of the input sequence into a Transformer-based policy.<n>We test our approach on a range of embodied Vision-and-Language tasks in our custom BabyAI-XGen environment.
  arXiv  Detail & Related papers  (2025-10-13T11:55:21Z)
• OmniQuality-R: Advancing Reward Models Through All-Encompassing Quality Assessment [55.59322229889159]
  We propose OmniQuality-R, a unified reward modeling framework that transforms multi-task quality reasoning into continuous and interpretable reward signals.<n>We use a reasoning-enhanced reward modeling dataset to form a reliable chain-of-thought dataset for supervised fine-tuning.<n>We evaluate OmniQuality-R on three key IQA tasks: aesthetic quality assessment, technical quality evaluation, and text-image alignment.
  arXiv  Detail & Related papers  (2025-10-12T13:46:28Z)
• Learning a Dense Reasoning Reward Model from Expert Demonstration via Inverse Reinforcement Learning [50.20267980386502]
  We learn a dense, token-level reward model for process supervision directly from expert demonstrations.<n>The learned reasoning reward serves two complementary roles: (i) it provides step-level feedback to optimise a reasoning policy during training; and (ii) it functions at inference as a critic to rerank sampled traces under fixed compute budgets.
  arXiv  Detail & Related papers  (2025-10-02T09:55:26Z)
• SOI is the Root of All Evil: Quantifying and Breaking Similar Object Interference in Single Object Tracking [25.076012214989433]
  We present the first systematic investigation and quantification of Similar Object Interference (SOI)<n> eliminating interference sources leads to substantial performance improvements (AUC gains up to 4.35) across all SOTA trackers.<n>We construct SOIBench-the first semantic cognitive guidance benchmark specifically targeting SOI challenges.
  arXiv  Detail & Related papers  (2025-08-13T06:12:43Z)
• Interpretable Reward Modeling with Active Concept Bottlenecks [54.00085739303773]
  We introduce Concept Bottleneck Reward Models (CB-RM), a reward modeling framework that enables interpretable preference learning.<n>Unlike standard RLHF methods that rely on opaque reward functions, CB-RM decomposes reward prediction into human-interpretable concepts.<n>We formalize an active learning strategy that dynamically acquires the most informative concept labels.
  arXiv  Detail & Related papers  (2025-07-07T06:26:04Z)
• An Active Inference perspective on Neurofeedback Training [0.6008132390640294]
  Neurofeedback training (NFT) aims to teach self-regulation of brain activity through real-time feedback.<n>NFT suffers from highly variable outcomes and poorly understood mechanisms, hampering its validation.<n>We propose a formal computational model of the NFT closed loop.
  arXiv  Detail & Related papers  (2025-05-06T08:41:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.