LFM2.5-2.6B, refusals removed
These are transformers weights, not a compiled model. There is no ExecuTorch
.ptein this repository: the one that used to sit inexecutorch/was built with a recipe that lost this family's tool calling and was withdrawn on 2026-09-19. Apps that run compiled exports, OpenWeights among them, cannot load safetensors. Use these withtransformers, or quantise them yourself; a replacement export needs its own GPTQ solve against these weights, which has not been done.
The ExecuTorch export was removed on 2026-09-19. It was built with a recipe carrying two defects: ExecuTorch 1.4.0's LFM2 definition never cleared
ShortConv.conv_statebetween prompts, so each prompt ran on the last one's state, and its int4 weights were round to nearest, which costs this family its tool calling. On 141 held-out questions the equivalent 1.2B export searched when needed on 10 percent of the questions that needed it and spoke of search results it had never fetched in a quarter of its replies; the solved re-export reads 49 and 1 percent at the same size and speed. Fixing it needs a GPTQ solve against these weights, which has not been done. The weights in this repository are unaffected. Method and numbers: https://github.com/alpharomercoma/openweights/blob/main/docs/research/executorch-state-and-recipes.md
Abliterated ("uncensored") version of LiquidAI/LFM2.5-2.6B, produced with
heretic: directional ablation of the refusal direction in the residual
stream, with a TPE search over per-layer ablation weights that minimises both the refusal rate and the
KL divergence from the original model. Weights are merged, so this loads as a plain Lfm2ForCausalLM.
Chat template, tool-calling tokens and everything else are unchanged from the base model.
heretic Pareto front for this run
study: ./checkpoints/LiquidAI--LFM2--5-2--6B.jsonl
completed trials: 200 / 200
idx trial refusals KL direction_index
0 171 0.03 0.1522 None
1 192 0.04 0.0806 None
2 191 0.05 0.0801 None
3 197 0.06 0.0648 None
4 8 0.07 0.0343 None
5 133 0.08 0.0238 17.379784537320543
6 92 0.11 0.0225 None
7 77 0.23 0.0210 None
8 158 0.24 0.0103 17.028759195631267
9 89 0.45 0.0051 17.485544563435877
10 45 0.79 0.0046 17.737500883050494
11 113 0.89 0.0035 16.83671987276088
12 194 0.96 0.0020 17.497441205715695
13 101 0.97 0.0016 18.074113894983057
14 97 0.98 0.0009 17.811708539059808
15 187 0.99 0.0008 11.863942700812128
selected pareto index: 4
ExecuTorch (CPU, KleidiAI on Arm)
| file | max context (tokens) | size |
|---|---|---|
executorch/lfm2_5_2_6b_heretic_8da4w_32k.pte | 32,768 | 1.81 GB |
Each is an ExecuTorch 1.4.1 program for the XNNPACK backend built with ExecuTorch's own LFM2 recipe: every linear
layer is quantised to 8-bit dynamic per-token activations and 4-bit grouped weights with group size 32 (8da4w,
the layout Arm's KleidiAI int4 kernels are written for), int8 embedding table, static KV cache, fused SDPA,
2048-token prefill chunk (longer prompts are prefilled in chunks; the context window is the figure in the table).
XNNPACK dispatches these matmuls to KleidiAI on every Arm CPU
(Apple Silicon, Snapdragon, Dimensity, Tensor, Graviton) in the stock ExecuTorch build; on x86 it falls back to AVX
kernels. The program is the same file on every platform.
The KV cache for the full window is allocated when the program loads (fp32, about 0.8 GB for 1.2B and 1.1 GB for
2.6B); the runtime attends over filled positions only, so the window costs memory, not speed. Run it with llama_main from ExecuTorch (examples/models/llama)
or the python runner, with this repo's tokenizer.json and the prompt format
<|startoftext|><|im_start|>user\n...<|im_end|>\n<|im_start|>assistant\n.
Pipeline
Reproducible end to end from https://github.com/ (see the repository this was built with):
make setup abliterate pareto, make save TRIAL=<idx>, make export, make publish.
Use responsibly; the safety training of the base model has been removed on purpose.