self-verification

[ICLR 2026] Official code for paper: TimeSearch-R: Adaptive Temporal Search for Long-Form Video Understanding via Self-Verification Reinforcement Learning.

Self-verification for LLMs.

Autonomous research system: self-directing LLM fleet with adversarial self-verification, on one workstation

Runtime reliability intelligence designed specifically for OpenClaw frameworks and agents.

入世（rushi-skill）：把书/课程/长内容蒸馏成可验证、可测试、可进化的 Agent Skill

LLM-as-a-Verifier (arXiv:2607.05391) as a dsh plugin — Best-of-N conversation mode: give DeepSeek V4 Flash test-time scaling. Bo5 self-verification hits 88% on Terminal-Bench 2.1, beating some frontier models at a fraction of the cost. Fine-grained logprob-expectation scoring, PPT tournament, zero-config.

把播客/长音频转写蒸馏成可验证、可测试、可进化的知识 Skill

LLM-as-a-Verifier plugin for DeepSeek Harness (DSH): best-of-N 候选并行生成 + 概率枢轴锦标赛(PPT)验证器择优, 官方样式设置卡片, lav_status/lav_set 会话工具 | DSH 插件: 让 AI 回答经过多候选验证器选优

Measure and prevent fake work in Claude Code sessions. 3-layer defense (memory feedback + Stop hook + slash command) + 4 audit tools. Baseline: 24.91% loose / 8.28% strict fake work rate across 813 real sessions.

Vision toolkit for AI agents — generate, analyze, and verify images. Local (Florence-2) + cloud (Gemini) vision with a see-verify loop for self-checking output.

Do Large-scale Language Models Eliminate the Need for Domain-Specificity in Automatic Term Extraction?

PhD Thesis

An n8n workflow that writes a daily Arabic tech digest and verifies its own output with an independent model before sending it.

Multi-source RAG system with self-verification, injection-safety, and PII checks for reliable, grounded answers.

KDD Cup 2026 DataAgent-Bench: ReAct agent for heterogeneous data analysis with fault-tolerant JSON parsing and SQL verification

LLM-Based Textual Verifier using Chain-of-Thought, Variant Generation, and Majority Voting.

VERDICT ANALYST, an agentic data analyst with a causal decision-science engine. Sandboxed code execution, self-verification, method routing (A/B, observational, time-series), and assumption diagnostics in plain business language.

Verify claims about your agent's own capabilities, not about the world.

Enterprise Agent RL Training & Self-Verification Platform - train agents with PPO/GRPO inside real harnesses, with LLM-as-a-Verifier self-verification

Purpose-aware company intelligence agent: give it a company and a reason (sales, competitor, interview prep) and it plans research around that reason, cites every claim, verifies its own work, and gets faster with every run. Hand-written state machine, memory, Reflexion-style lessons.