[KongchangAI]
ModelDiSPaT框架

DiSPaT

A tabular data anomaly detection framework based on self-play fine-tuning. It uses an LLM to serialize table rows into text and alternately optimizes a policy model and a discriminator under an f-divergence minimization framework, tightening the characterization of the normal distribution without requiring real anomaly labels.

Timeline (last 90 days)

Nov 16

DiSPaT is a method that strengthens a model's understanding of normal data through a self-play fine-tuning framework, for tabular anomaly detection.

Unverified50%
Nov 16

DiSPaT is based on a theoretical framework of f-divergence minimization and derives a tight approximation, whose training objective is equivalent to narrowing the gap between the true normal data distribution and the distribution of samples generated by the model.

Unverified50%
Nov 16

DiSPaT uses an alternating optimization mechanism: the policy model generates pseudo-anomalous samples, the discriminator learns to distinguish synthetic samples from real normal samples, and the discriminative signal is fed back to drive policy updates to align the distributions.

Unverified50%
Nov 16

DiSPaT's training process does not require real anomaly labels; the model generates pseudo-anomalies itself and then learns to identify them, thereby continuously tightening its definition of normality.

Unverified50%
Nov 16

DiSPaT shares ideas with generative adversarial networks, but its theoretical foundation is built on f-divergence minimization rather than JS divergence.

Unverified50%
Nov 16

Experiments on diverse benchmarks show that DiSPaT consistently outperforms prior LLM-based methods, deep learning anomaly detection methods, and classic unsupervised detectors.

Unverified50%
Nov 16

The computational cost of self-play training is higher than one-off fine-tuning, because each iteration must generate samples and train the discriminator.

Unverified50%
Nov 16

The paper does not specify how pseudo-anomalies are generated, nor does it give the number of rounds and cost required for convergence.

Unverified50%
Nov 16

For tabular-data-intensive scenarios such as industrial quality inspection, financial risk control, and log auditing, DiSPaT is expected to reduce reliance on anomaly annotations.

Unverified50%

All Facts (9)

Unverified

DiSPaT is a method that strengthens a model's understanding of normal data through a self-play fine-tuning framework, for tabular anomaly detection.

50%
Unverified

DiSPaT is based on a theoretical framework of f-divergence minimization and derives a tight approximation, whose training objective is equivalent to narrowing the gap between the true normal data distribution and the distribution of samples generated by the model.

50%
Unverified

DiSPaT uses an alternating optimization mechanism: the policy model generates pseudo-anomalous samples, the discriminator learns to distinguish synthetic samples from real normal samples, and the discriminative signal is fed back to drive policy updates to align the distributions.

50%
Unverified

DiSPaT's training process does not require real anomaly labels; the model generates pseudo-anomalies itself and then learns to identify them, thereby continuously tightening its definition of normality.

50%
Unverified

DiSPaT shares ideas with generative adversarial networks, but its theoretical foundation is built on f-divergence minimization rather than JS divergence.

50%
Unverified

Experiments on diverse benchmarks show that DiSPaT consistently outperforms prior LLM-based methods, deep learning anomaly detection methods, and classic unsupervised detectors.

50%
Unverified

The computational cost of self-play training is higher than one-off fine-tuning, because each iteration must generate samples and train the discriminator.

50%
Unverified

The paper does not specify how pseudo-anomalies are generated, nor does it give the number of rounds and cost required for convergence.

50%
Unverified

For tabular-data-intensive scenarios such as industrial quality inspection, financial risk control, and log auditing, DiSPaT is expected to reduce reliance on anomaly annotations.

50%

Source Articles