Generative Adversarial Loops

2026-10-08

AI Summary

This study addresses the lag of existing AI benchmark construction behind methodological innovation and the absence of autonomous evolution mechanisms. To this end, it proposes a generative adversarial loop framework that introduces a discriminator agent for automated objective setting to generate adversarial data exposing model weaknesses, thereby driving generator algorithms to iteratively overcome these deficiencies and achieve closed-loop self-evolution from weakness discovery to algorithmic improvement. By integrating Generative Adversarial Network principles with agent-based search techniques, this framework is applicable to efficient inference scenarios such as KV cache compression, sparse video processing, and attention mechanisms. Experimental results demonstrate substantial performance improvements across multiple tasks, notably elevating the KV compression score from 0.35 to 0.97, while surpassing current state-of-the-art methods on standard benchmarks.

Abstract

AI research progress can be viewed as the interaction between two processes: benchmark creation and method discovery. Historically, both were driven by human intelligence. However, recent advances in AI have accelerated automated method discovery, while automated benchmark creation has received comparatively less attention. To enable self-advancing systems, we propose Generative Adversarial Loop (GAL), a generator-discriminator framework alternating between two agentic searches: (1) a discriminator that generates adversarial data to expose weaknesses in current systems, and (2) a generator that discovers algorithms to overcome them. We apply this framework to approximation algorithms for efficient inference. Unlike existing auto research systems, which primarily focus on algorithm discovery, GAL introduces a discriminator agent that automates goalpost setting by continually searching for weaknesses in the current algorithm. We demonstrate adversarial data generation across four tasks: KV compression, sparse video generation, sparse attention, and context extension, where the discriminator identifies weaknesses in state of the art techniques. We further show that GAL enables autonomous improvement, with newly discovered algorithms improving not only on adversarially generated data, but also on established benchmarks. Specifically, GAL improves CompactorPress on KV compression with Qwen3-4B at 4x, raising performance on the discriminator dataset from 0.35 to 0.97, while also outperforming RULER-HARD (+0.77 pts). For context extension, GAL boosts Dual Chunk Attention from 0.20 to 0.90 on the discriminator dataset, while yielding gains on standard benchmarks(ScienceFiction (+6 pts) and PG19 32K (-0.33 PPL)). GAL thus provides a path toward autonomous goalpost setting and algorithmic improvement, where AI systems continually discover their own weaknesses and develop methods to overcome them.

Problem

• Automated Benchmark Creation
• Autonomous Goalpost Setting
• Self-advancing AI Systems
• Adversarial Data Generation
• Weakness Discovery

Innovation

• Generative Adversarial Loop
• Automated Benchmark Creation
• Discriminator Agent
• Autonomous Improvement
• Approximation Algorithms