Revisiting Long-context Modeling from Context Denoising Perspective
This addresses a key bottleneck in long-context modeling for applications requiring robust sequence processing, though it is incremental as it builds on existing noise analysis.
The paper tackles the problem of long-context models being misled by irrelevant tokens (contextual noise) by proposing a metric to detect noise and a training strategy to mitigate it, resulting in an 8B model achieving performance (50.92) comparable to GPT-4o (51.00).
Long-context models (LCMs) have demonstrated great potential in processing long sequences, facilitating many real-world applications. The success of LCMs can be attributed to their ability to locate implicit critical information within the context for further prediction. However, recent research reveals that LCMs are often susceptible to contextual noise, i.e., irrelevant tokens, that can mislead model attention. In this paper, we conduct a fine-grained analysis of the context noise and propose an effective metric, the Integrated Gradient (IG) score, to detect and quantify the noise information within the context. Our findings reveal that even simple mitigation of detected context noise can substantially boost the model's attention on critical tokens and benefit subsequent predictions. Building on this insight, we propose Context Denoising Training (CDT), a straightforward yet effective training strategy that improves attention on critical tokens while reinforcing their influence on model predictions. Extensive experiments across four tasks, under both context window scaling and long-context alignment settings, demonstrate the superiority of CDT. Notably, when trained with CDT, an open-source 8B model can achieve performance (50.92) comparable to GPT-4o (51.00).