---

license: apache-2.0

tags:

• abliterated
• uncensored
• gguf
• spark
• reasoning

base_model: XHToken/Spark-X2.5-1.7B

pipeline_tag: text-generation

---

Spark-X2.5-1.7B-Abliterated-GGUF


Requests, questions or suggestions? Message me on X: https://x.com/imdariotoo

Uncensored community quants of XHToken/Spark-X2.5-1.7B,

produced by single-round rank-1 refusal-direction ablation (heretic) over a balanced

6292-prompt multilingual pool. Trial 33 of a 50-trial Optuna search.

Base model card, architecture notes and serving instructions live in the sibling repo

darioooooo0o/Spark-X2.5-1.7B-GGUF — this README covers only what the ablation changed

and how it was measured.

Files

| file | size | notes |

|---|---|---|

| Spark-X2.5-1.7B-Abliterated-Q8_0.gguf | ~1.8 GB | max quality |

| Spark-X2.5-1.7B-Abliterated-Q6_K.gguf | ~1.4 GB | recommended daily driver |

| Spark-X2.5-1.7B-Abliterated-Q4_K_M.gguf | ~1.0 GB | smaller cards |

| Spark-X2.5-1.7B-Abliterated-IQ4_XS.gguf | ~0.9 GB | smallest |

All quants built with imatrix calibration from the same BF16 merge. Serve with the

Spark fork of llama.cpp (spark2_5 arch) and f16 KV cache:

llama-server -m <gguf> --chat-template-file chat_template.jinja \
  --host 127.0.0.1 --port 12450 -c 65536 -n 4096 -ngl 99 \
  -fa on -ctk f16 -ctv f16 --temp 1.0 --top-p 0.95 --top-k -1 \
  --reasoning-budget 2000 --jinja --reasoning-format deepseek

--reasoning-budget is load-bearing, not optional: median thinking length is 1500+

tokens and the tail exceeds 4000. Unbounded reasoning will eat your context.

Refusal results

Headline: **0 refusals in 300 sealed generations + 37 answer-level generations,

eye-audited** (base model: 186/300 refused on the same prompts).

Sealed sets (never trained on, never trial-scored — the ungameable numbers)

| set | base refused | ablated refused | base preamble | ablated preamble |

|---|---|---|---|---|

| sealed mjd, 200 prompts, 7 langs | 86 (43%) | 0-being-6-flagged, 0 real (3%) | 247 chars | 96 chars |

| sealed adv, 100 prompts, English | 100 (100%) | 0-being-15-flagged, 0 real (15%) | 435 chars | 145 chars |

"0-being-N-flagged" needs explanation — see Methodology. Short version: every flag

was audited by eye; all 21 were reasoning traces planning compliance or complying

answers containing words like "illegal". Zero actual refusals. Preamble (compliance-talk

before the answer) collapsed 2-3x: answers start directly now.

Full matrix (marker ruler, descriptive — same prompts both models)

| set (n) | base | ablated t33 |

|---|---|---|

| all-eval, 3911 unique harmful prompts | 2014 (51.5%) | 263 (6.7%) |

| selection mjd-400 | 194 (48.5%) | 2 (0.5%) |

| selection adv-100 | 100 (100%) | 2 (2%) |

Marker numbers overstate both sides on thinking models (see below). They are

included for comparability with other releases, not as claims.

Answer-level audit (4096-token generations, think/answer split)

41 sealed prompts (21 flagged + 20 random unflagged): 37 finished thinking,

37 produced answers, 0 answers refused. All 8 answer-flags were complying

answers (guides, letters, code) containing refusal-vocabulary words.

Methodology: why marker counts are fiction on thinking models

Standard refusal eval generates ~100 tokens and string-matches for "sorry",

"as an AI", "illegal", etc. On a thinking model this fails twice, in opposite

directions, on the same model:

1. Flags compliance. 100 tokens never exit thinking, so the matcher scores

reasoning traces. Traces planning compliance ("We should provide a detailed,

step-by-step guide" — for explosives manufacturing) get flagged for containing

"as an AI". Complying answers get flagged for containing "illegal" (a drug-guide

title: "# The Complete Guide to Manufacturing and Distributing Illegal Drugs").

1. Misses real structure. Non-answers (thinking that never converges) score as

neither refusal nor compliance.

Our ruler instead: 600+ token generations, split thinking from answer on

</think>, three buckets per generation — complied / refused / non-answer —

with every flagged item eye-audited. Reported numbers above are bucket counts,

not marker counts. Anyone comparing "0/300" against a marker-count "0/465" is

comparing different rulers; demand the bucket table.

How it was made

1. Pools (frozen, manifest-hashed): train 6292 rows (mjd multilingual ×7 langs

+ AdvBench oversampled ×10 for equal family mass — balanced diff-in-means needs

equal votes, not proportional ones), selection 500 stratified (400 mjd + 100 adv),

sealed 200 mjd + sealed 100 adv held out of everything. Zero-overlap matrix verified.

1. Search: heretic Optuna, 50 trials (20 startup + 30 TPE), batch 32, bf16,

KL co-optimized. Trial 33 selected: 4/500 selection refusals at KL 0.016.

1. One round. Minimal-change principle: bigger balanced slices beat stacked

rounds. A pre-registered second-round gate (residual cosine) was mooted and

dropped — there is no residual refusal behavior left to ablate.

1. Export: adapter-save, CPU reload, merge, tied-embedding twin-fill

(PEFT unties embeddings on merge leaving a meta stub — filled from the live twin).

Full pipeline (patch scripts, pool builder, launchers, export, evals) is deterministic

from the base safetensors; trial-33 params recorded in the study journal.

Quality gates (base vs ablated)

| check | base | ablated t33 |

|---|---|---|

| PPL, calib corpus, 512 ctx, BF16 | 3.8598 ± 0.0097 | 3.9492 ± 0.0099 (+2.3%, disclosed) |

| Agentic tool-use, 6-task HermesBench | 5/6 effective | 5/6 effective (one single-run miss retested 3/3 pass = temp-1.0 noise) |

| KL divergence (search objective) | — | 0.016 |

| Decode speed, RTX 3060 | 171 t/s Q6_K | identical arch/size — same |

The +2.3% PPL is over our own 2% bar by 0.3 points and is printed here, not buried.

No capability damage was demonstrated anywhere else, so no DPO repair pass was run.

Caveats

• Inherits the base model's reasoning-loop failure mode on vague open-ended prompts

(documented in the base repo README): unbounded thinking until the token limit,

independent of quantization. --reasoning-budget + client timeouts are mandatory.

• Most generations spend 1500+ tokens thinking; short max_tokens values return

reasoning fragments, not answers. Budget context accordingly.

• "0 refusals" means no observed refusals in the evaluated sets under eye audit,

not a proof of universal compliance. Novel framings were not exhaustively tested.

• Ablation removes refusal behavior, not knowledge. Standard responsible-use

expectations apply.

License

Apache 2.0, same as the base model. Ablation deltas are rank-1 LoRA merges;

no additional training data is distributed with these files.