Loading
Life sciences · Preprint
arXiv · September 5, 2026
Early or partial results. Treat as a signal, not a conclusion.
Cadence is a novel lossy compression system combining a 330M-parameter time-series foundation model (TimesFM-3) with adaptive arithmetic coding, designed to bound error on every compressed sample. Testing on electricity demand and transit ridership data from 2026 shows improvements over classical predictors and downsampling, but the work is unpublished, unvalidated in clinical or medical contexts, and the authors explicitly document failed generalization to external benchmarks.
Algorithm development and empirical comparison on real-world time series. 49 electricity demand time series from EIA-930 balancing authorities (2026); 50 transit ridership series from MTA (2026); 27 series from SDRBench external benchmark. All operational/utility datasets; no medical or clinical population.. Intervention: Cadence error-bounded lossy compressor (TimesFM-3 foundation model + adaptive arithmetic coder with context-modelled binarization). Compared with: Six classical predictors (unspecified); xz/zstd for index compression; downsampling (standard database retention); SDRBench external benchmark.
Against six classical predictors, Cadence gains 13.3% on 49 EIA-930 balancing-authority demand series and 28.3% on 50 MTA ridership series. Median gain across 297 series-tolerance pairs is 21.4%, winning all 297 pairs tested. Against downsampling (standard database retention method), guaranteed worst-case error is 28–56× tighter at equal size.
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
A technical systems paper describing a novel compression algorithm paired with a foundation model, demonstrating proof-of-concept on real time-series data but lacking clinical validation, peer review, or established relevance to clinical decision-making.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
We present Cadence, an error-bounded lossy compressor for numeric time series pairing a 330M-parameter time-series foundation model (Google TimesFM-3) with an adaptive arithmetic coder, guaranteeing $|\hat{x}_t-x_t|\leτ$ on every sample. One negative result constrains the design space: for lossless coding a foundation model is worth nothing, because bits saved are logarithmic in predictor accuracy, $Δb=\log_2(\mathrm{MAE_{old}}/\mathrm{MAE_{new}})$. So the $1.51\times$ advantage TimesFM-3 holds over a 32-tap linear predictor buys 0.60 bits of 20.28, a median gain of +0.03%. Error-bounded coding escapes this at one point: once a forecast lands inside the band the residual index is zero and the sample nearly free. Cadence contributes: (1) an adaptive range coder with context-modelled binarization, beating xz/zstd on real indices by 9.7% (15/15) and reversing a finding from a general-purpose back end; (2) a determinism result -- predictions are not bit-identical across batch sizes, and no PyTorch configuration repairs this, forcing group size and execution device into the container format; and (3) domain localization on corpora postdating any plausible training cutoff. On 49 EIA-930 balancing-authority demand series (2026) Cadence gains 13.3% over the best of six classical predictors, and 28.3% on 50 MTA ridership series (2026): 21.4% median over 297 series-tolerance pairs, winning all 297. Against downsampling, what time-series databases deploy for retention, its guaranteed worst-case error is $28$--$56\times$ tighter at equal size. End-to-end, once the context bootstrap is paid for, gains run from 6.8% at six months of hourly data to 15.1% asymptotically. Attempting to falsify the domain claim on SDRBench, theory predicts failure and delivers: -0.8% median, 0 of 27 pairs gaining. Three further negative results and eight retracted claims are reported in full.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.