MT-Bench
Harbor: pendingWe explore the use of LLM-as-a-judge for evaluating chat assistants and show it matches human preferences with an agreement over 80%.
- Samples
- N/A
- Released
- 2023
- Scores
- 0
Perceive, Solve, Reason Suite
Perceive, Solve, Reason: Progressive Capability Formation in Time Series-LLMs
A suite with 337K training and 44K evaluation samples measuring perception, solving, and reasoning capabilities for Time Series LLMs across synthetic and real-world domains.
How to run
ltb · perceive-solve-reason-suite
$ npm i -g lettokenburn-cli$ ltb bench info perceive-solve-reason-suite
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
2 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Qwen3 1.7B | 79.0 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
| Qwen3 4B | 0.6 | N/A | Independent | medium | Oct 16, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
The suite curates four public corpora into three progressive stages: Perceive (attributes), Solve (forecasting/anomaly detection), and Reason (explanation/decision). It includes held-out evaluation sets and reserves three external benchmarks for out-of-domain testing.
Scoring compares encoder-based designs against textualization, highlighting that dedicated encoders and progressive training order significantly improve reasoning accuracy. The benchmark evaluates model performance on perception attributes and downstream tasks.