LFM2-350M
LiquidAI/LFM2-350M
0.4BLFM6 Heretic artifacts4 measured conversions3.8K downloads
Best version95.0Pareto
MihaiPopa-1/LFM2-350M-heretic-safety-defiltered-high-reasoningby MihaiPopa-1 · Heretic v1.1.0 · SAFETENSORS
- KL divergence
- 0.042
- Refusals
- 6/100
- Refusal rate
- 6%
- Downloads
- 114
Evidence from the model card
| KL divergence | 0.0420 | | Refusals | 6/100 |
Metrics published by the author. Why this ranks first.
All measured versions
Higher score is better. One row per ablation attempt; artifacts of the same attempt are grouped under it.
| # | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|
| 1 | MihaiPopa-1 | 0.4B | 0.042 | 6/1006% | 95.0 | BestParetoSAFETENSORS | 114 |
| 2 | MihaiPopa-1 | 0.4B | 0.021 | 10/10010% | 94.0 | ParetoSAFETENSORS | 123 |
| 3 | MihaiPopa-1 | 0.4B | 0.084 | 5/1005% | 93.6 | ParetoSAFETENSORS | 138 |
| 4 | marduk191 | 0.4B | 0.137 | 12/10012% | 88.0 | GGUFSAFETENSORS | 3.4K |
All artifacts (6)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| LFM2-350M-heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.137 | 2.3K | 2025-12-29 |
| LFM2-350M-heretic-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.137 | 853 | 2025-12-29 |
| LFM2-350M-heretic marduk191 | heretic | SAFETENSORS | published (weak match)KL 0.137 | 258 | 2025-12-29 |
| LFM2-350M-heretic-safety-defiltered MihaiPopa-1 | heretic | SAFETENSORS | published (weak match)KL 0.084 | 138 | 2026-02-12 |
| LFM2-350M-heretic-safety-defiltered-xhigh-reasoning MihaiPopa-1 | heretic | SAFETENSORS | published (weak match)KL 0.021 | 123 | 2026-02-11 |
| LFM2-350M-heretic-safety-defiltered-high-reasoning MihaiPopa-1 | heretic | SAFETENSORS | published (weak match)KL 0.042 | 114 | 2026-02-11 |