Papersby GlyphVeriHarness: Scaling Agentic Verification for Long-Horizon Tasks research paper by Google, 2026
Google · Alignment and safety · Oct 1, 2026 · 1 citations · 62 upvotes · unverified
What it shows
VeriHarness improves long-horizon agent performance by using the base model as a verifier to resolve conflicting claims and challenge consensus.
By Caiqi Zhang, Rujun Han, Zifeng Wang and 4 more · arXiv 2610.00972 · PDF · Code
UnverifiedSummarised by a model from the abstract; not yet checked by hand.
More from Google
All 45EmbeddingGemma 2: Unified and Efficient Multimodal RepresentationsEmbeddingGemma 2 is an open multimodal embedding model that unifies text, images, audio, and video representations with efficient resource-flexible design and state-of-the-art benchmark performance.
Google34 upvotesRead →StreamVibe: Proactive Video Agents for Evolving User IntentStreamVibe introduces a benchmark, streaming vision-language model, and hierarchical memory harness to enable proactive, intent-aware interaction with implicit and evolving user goals.Google78 upvotesRead →On-Policy Self-Training: Data-Free Self-Improvement in Large Language ModelsFoundation models can self-improve across text and vision tasks by generating synthetic questions and reasoning traces guided only by high-level topics, without external supervision.Google31 upvotesRead →DeLoop: Repeated Actions as Self-Supervision for Long-Horizon AgentsThis self-supervised training method teaches AI agents to avoid repetitive, unproductive actions during long tasks by using their own history as feedback.Google21 upvotesRead →CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation PipelineThis paper introduces a benchmark and generation pipeline to evaluate how agents resolve conflicts and verify information in multi-agent computer-use scenarios.Google36 upvotesRead →Harness^2: Recursive Agent Harnessing for an Open WorldThis paper introduces a recursive framework that allows agents to automatically improve their own execution harnesses for better performance on novel open-world tasks.Google2 citations123 upvotesRead →VIS-Ground: Video Interactive Storytelling with Contextual GroundingThis paper introduces a framework that transforms heterogeneous user inputs into structured constraints to improve grounding in interactive video generation.Google103 upvotesRead →SEER: Self-Evolving Event Reasoning and Retrieval for Time Series ForecastingSEER improves time series forecasting by using a closed-loop system to retrieve and reason about external events that influence data trends.Google35 upvotesRead →
Other alignment and safety papers
TopicTraining language models to follow instructions with human feedbackInstructGPT: fine-tuning on human feedback made a 1.3B model preferred over the 175B GPT-3.
OpenAI25.8k citations26 upvotesRead →Direct Preference Optimization: Your Language Model is Secretly a Reward ModelDPO aligns a model to human preferences with a simple classification loss, no reward model or RL loop.
Stanford University11.6k citations73 upvotesRead →Constitutional AI: Harmlessness from AI FeedbackConstitutional AI trains a harmless assistant from AI feedback guided by a short list of written principles, not human harm labels.
Anthropic3,987 citations4 upvotesRead →Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMsReinforcement Learning from Human Feedback (RLHF) benefits from simpler REINFORCE-style optimization methods over computationally expensive PPO or "RL-free" methods, enhancing performance and reducing cost in large language models.
Cohere804 citations18 upvotesRead →Sleeper Agents: Training Deceptive LLMs that Persist Through Safety TrainingModels trained with a hidden backdoor kept their deceptive behaviour through standard safety training.Anthropic657 citations30 upvotesRead →Alignment faking in large language modelsClaude 3 Opus sometimes went along with a training goal it disagreed with, to avoid being changed, without being told to.Anthropic393 citations10 upvotesRead →
About this paper
- Citations
- 1, 0 influential · Semantic Scholar
- Upvotes
- 62 · Hugging Face
- Lab
- Google · on Companies · on Acquisitions · on Paydays · on TechConf · on Releases
Changes
A first count of 0 influential citations.
A first count of 1 citations.
VeriHarness improves long-horizon agent performance by using the base model as a verifier to resolve conflicting claims and challenge consensus.
Moved from Agents and evaluation to Alignment and safety.
Found by the weekly Hugging Face scan; not yet checked by hand.