Is NTK-aware superseded?

NTK-aware (Long-context / context-window extension): superseded — cited as a baseline and beaten by newer methods. 4 paper(s) critique it, 6 beat it on benchmarks — #6 of 53 most-superseded. Sub-problem: cluster led by YaRN. Newer alternatives in the same sub-problem include Cross-Resolution Phase-Aligned Attention (CRPA), DoPE, DyPE.

Long-context / context-window extension

NTK-aware

Superseded baseline#6 of 53 most-superseded

Superseded — cited as a baseline and beaten by newer methods

4 papers critique it · 6 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites NTK-aware as a baseline.

rescaling factors derived from previous methods often fall short of achieving the effective target context length.

— LongRoPE2: Near-Lossless LLM Context Window Scaling

“PE and NTK experience repetition issues, resulting in lower NoRepeat Score.”

— RIFLEx: A Free Lunch for Length Extrapolation in Video Diffusion Transformers

“However, these static approaches do not account for the distinctive spectral progression of the diffusion process, where low-frequency structures are generated in the first sampling steps, while high-frequency details are resolved later”

— DyPE: Dynamic Position Extrapolation for Ultra High Resolution Diffusion

“methods like NTK, Dyn-NTK, and YaRN suffer from attention logit outliers due to their positional embedding interpolations”

— A Training-Free Length Extrapolation Approach for LLMs: Greedy Attention Logit Interpolation (GALI)

Beaten on benchmarks

Head-to-head results where a newer method reports beating NTK-aware. Values are copied from the source paper's tables — verify against the cited paper.

TokenSelect beats NTK-aware
36.77 vs 0.83
Avg. · [Yi-1.5-6B]
TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection
ParallelComp beats NTK-aware
51.27 vs 5.35
Average · [Llama2-7B-chat-hf(4k)]
ParallelComp: Parallel Long-Context Compressor for Length Extrapolation
RULER beats NTK-aware
23.00 vs 7.57
LOFT Avg. · [Base model: Phi3-mini (3.8B)]
LongRoPE2: Near-Lossless LLM Context Window Scaling
Cross-Resolution Phase-Aligned Attention (CRPA) beats NTK-aware
75.34 vs 44.52
Overall (All) · [video generation with DOVE metrics]
One Attention, One Scale: Phase-Aligned Rotary Positional Embeddings for Mixed-Resolution Diffusion Transformer
DoPE-by-Gaussian beats NTK-aware
70.083 vs 60.938
Original (64k) · [64k tokens]
DoPE: Denoising Rotary Position Embedding

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.