precise-anomaly-localization-long-multivariate-time-series-16k-131k-points-keeping CLM-0626For precise anomaly localization in long multivariate time-series (16k-131k points), keeping the signal outside the model context as a variable the model queries through code over multiple turns recovers far more localization accuracy than single-pass ingestion, where the recording is serialized, subsampled or compressed into context once.
Capability: Losing information in long inputs
Observed on
Synthetic benchmark with programmatically injected parametric anomalies plus ECG/sleep holdouts; frontier backbones (GPT-5.5) benefit most, 4B open backbones need RL post-training;.
Sources
- supportsTimeRLM: Recursive Language Models Enable Precise Anomaly Localization in Long-Context Time-SeriesValidated by: On AnomalyXL-Localize, IoU 0.682 vs at most 0.329 for all single-pass LLM, TSLM and fine-tuned 32B Toto baselines; classify-with-evidence 0.745 vs below 0.08; zero-shot ECG/sleep IoU 0.271 vs under 0.004 for every TSLM. Scalar-magnitude task did not improve (0.440 to 0.480).. Measured on a self-introduced synthetic benchmark (200 localize samples, 100 for multichannel tasks) plus 250+250 real clinical windows; baselines include three fine-tuned TSLMs, a classical feature+GBT model and a 32B TSFM-VLM. [truncated]
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- On long-video (10-150 minute) audio-visual reasoning questions, multiple-choice accuracy overstates grounding: a large open model can answer about 55% of items correctly while scoring 8.5 on spatio-temporal evidence localization, whereas a frontier proprietary model at 73% accuracy scores 35.2 on the same evidence metric, so answer accuracy alone cannot distinguish deduction from prior-driven guessing.Reasoning about time in video · reviewed by AI
- A linear probe on a tool-calling model's hidden state at the last generated token detects incorrect tool-calls — including wrong-but-well-typed argument values that parsers and logs do not catch — and works better on larger models and at middle-to-late layers than at the final layer.Using the tools it is given · reviewed by AI
- On SWE-bench Lite, asking the model to name suspicious files from the issue plus a repository tree finds the ground-truth file in about 79% of cases and embedding retrieval over the issue text in about 68%, and taking the union of the two raises file-level localization to about 82% because the two methods miss different files; a fixed three-step narrowing (files, then classes and functions, then edit lines) with no agent loop then supported the best open-source resolve rate at the time (32%).Finding what matters in a large codebase · reviewed by AI
- In training-free video temporal grounding with frozen vision-language embedding backbones, matching a query only against frames sampled at a fixed rate leaves temporal context unused: adding clip-level (sliding-window) similarities projected onto the frame timeline raises localization accuracy even though clip-only similarity is worse than frame-only on its own.Reasoning about time in video · reviewed by AI
- A linear probe on a tool-calling model's hidden state at the last generated token detects incorrect tool-calls — including wrong-but-well-typed argument values that parsers and logs do not catch — and works better on larger models and at middle-to-late layers than at the final layer.Using the tools it is given · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: A single-pass model that ingests the full uncompressed recording matching recursive code-querying IoU on span-localization tasks at comparable length, or recursive querying failing to beat single-pass on a non-synthetic long-context localization set.