Ulisse E4B: gemma-4-E4B-it, think symbolic image

Ulisse E4B is the effective-4-billion-parameter member of the Ulisse line, the thinking lookers' second family beside Lucia, on Gemma 4: a LoRA adapter of google/gemma-4-E4B-it that looks at a picture and writes the row of thought-symbolic-image-captions: a short chain of plain-language thoughts, one named step per line, then the wordings (a caption, a prompt, tags, a critique, by depth), then the questions a careful looker would still ask, each with its answer and what settles it. Every size is taught the same rows under the same recipe; the sizes differ in what they can hold, and the reads below say what this one holds. He writes the row's form whole and backs his thoughts at the Lucia 4B's level; the print term and the long specification rows are where he falls short of the Qwen sizes. The fastest looker of the group.

see: A single figure, seated, holding a book, in three-quarter view.
place: The figure sits at the centre, the book held low against the lap.
form: The sitter fills the frame from the knees up, the background a plain dark wash.
colour: Brown ink over graphite, the paper a warm cream.
mood: The lowered eyes and the still hands suggest absorption, a reader lost in the page.
caption: A seated figure reads a book held in the lap, drawn in brown ink over graphite.
prompt: drawing, seated figure reading, three-quarter view, brown ink over graphite, cream paper, plain dark wash
question: thing name -> book [record]
question: medium -> ink and graphite [record]
question: mood tone -> absorbed [chain]
question: sitter identity -> unknown [unknown]

The steps a chain may use, in the order a careful looker takes them: see, name, place, count, relate, form, light, colour, mood, mean, intend, story, compare, change, check, say. A quick look holds 1 to 2 thoughts, a closer look 3 to 5, a long look 6 to 10. Questions end in [record] (the answer is a fact of the picture), [chain] (the chain worked it out) or [unknown]. The prompt line's first term names the kind of picture (painting, print, drawing, photograph).

Loading

This repository holds the LoRA adapter; the Gemma 4 line ships adapters, no full fine-tune. Every Ulisse loads the same way and is run the same way: thinking off through the template's own switch, greedy decoding, no system prompt (the row's form is learned). Needs transformers 5.10 or newer (Gemma 4) and peft 0.18 or newer.

The LoRA adapter (AbstractPhil/gemma-4-e4b-think-symbolic-image): the base model, the adapter merged into it, the processor from the adapter repository (the base processor's files are kept there).

import torch
from peft import PeftModel
from transformers import AutoModelForMultimodalLM, AutoProcessor

processor = AutoProcessor.from_pretrained("AbstractPhil/gemma-4-e4b-think-symbolic-image")
model = AutoModelForMultimodalLM.from_pretrained("google/gemma-4-E4B-it", dtype=torch.bfloat16, device_map="cuda")
model = PeftModel.from_pretrained(model, "AbstractPhil/gemma-4-e4b-think-symbolic-image").merge_and_unload().eval()

The full fine-tune: none for the Ulisse line; the adapter is the release here.

Then:

from PIL import Image

image = Image.open("picture.jpg").convert("RGB")
prompt = ("Look at this picture and think it through in writing, a closer look, ending with a caption; then write a caption "
          "and a prompt or tags, and the questions a careful looker would still ask.")
messages = [{"role": "user", "content": [{"type": "image", "image": image}, {"type": "text", "text": prompt}]}]
text = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=False, enable_thinking=False)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=800, do_sample=False)
print(processor.batch_decode(out[:, inputs["input_ids"].shape[1]:], skip_special_tokens=True)[0])

The depth words: "a quick look" (1 to 2 thoughts), "a closer look" (3 to 5), "a long look" (6 to 10). The parser, the scorer and the trainer are in qwen-symbolic-thinking (qwen_symbolic_thinking.chain.parse, chain.score); the same package extracted every number below.

The reads

Read the same way as the Lucia family: 60 pictures and 40 specifications no training saw, against the dataset's own rows (row shape, said / backed, definitions); then the medium term on two sets no training row came from: 500 National Gallery of Art pictures (100 drawings, prints, paintings, photographs and objects each) and 200 pictures of AIC, NGA, Cleveland and the Met. Lucia 4B (the Qwen3.5 member of the same class) and Lucia v1 (the 9B) stand beside.

read	Ulisse E4B	Lucia 4B	Lucia v1 (9B)
row shape, pictures / specifications	1.00 / 1.00	1.00 / 1.00	1.00 / 1.00
said / backed, whole rows	0.17 / 0.28	0.19 / 0.26	0.19 / 0.30
definitions, pictures / specifications	0.00 / 0.28	0.02 / 0.12	0.02 / 0.08
500 held-out pictures: painting / print / drawing / photograph / object named rightly of 100	83 / 71 / 80 / 82 / 90	83 / 78 / 83 / 81 / 91	83 / 76 / 86 / 80 / 90
200 museum pictures: painting / print / drawing / photograph / object	13 of 27 / 36 of 39 / 12 of 13 / 12 of 13 / 66 of 66	14 of 27 / 36 of 39 / 12 of 13 / 11 of 13 / 66 of 66	13 of 27 / 38 of 39 / 12 of 13 / 11 of 13 / 66 of 66
the 200 museum pictures on one RTX PRO 6000, batch 64, 800 new tokens a row	0.22 s a row	0.26 s a row	0.27 s a row

What this says: the Ulisse E4B holds the row's form whole and backs what it says at the Lucia 4B's level (0.28 against 0.26 on the 60-row screen; that read moves by 0.03 to 0.05 between seeds of one recipe), while saying a little less of a picture's caption (0.17 against 0.19). On the medium term it names paintings, photographs and objects as the Qwen sizes do and falls short on prints (71 of 100 against 76 to 78) and a little on drawings (80 against 83 to 86); the museum pictures sit at the family's level. Where it differs most is the long specification rows, written from a text description alone: its thoughts there carry a defining phrase in 28% of rows against the Qwen sizes' 8 to 12%, and its stock mood words come a little more often (48% of the 500 held-out rows against 44%). It is the fastest looker of the group at 280 image tokens a picture. Before training, the base model already wrote the row's shape from the plain ask (format 1.00, 65% of rows at their depth, 4.5 thoughts), which no Qwen base did; the training taught it the backing and the house vocabulary. The museum painting row understates every model: watercolours and gouaches on paper are named drawings, as the National Gallery's own catalogue does. The reads are made with qwen_symbolic_thinking.tools (probe_read, museum_read, museum_mismatch, screen_conflicts), so any arm can be read the same way.

How he was trained


base	google/gemma-4-E4B-it, bf16, the vision encoder frozen, 280 image tokens a picture (the model's default budget)
adapter	LoRA rank 32, alpha 64, dropout 0.05 on the language model's q, k, v, o, gate, up and down projections of all 42 layers (the key and value projections exist on 24 of them; 69.8 M trainable parameters, saved in fp32)
data	Lucia v1's rows exactly: 951 picture rows taught twice, 500 long-look drawn rows, 2,000 synthetic closer looks and 17,385 museum rows whose catalogue gives the medium, the prompt line's first term set to it (6,606 drawings, 3,523 paintings, 3,000 prints, 2,016 photographs, 2,240 objects); 20,836 items, 21,787 examples; the 500 probe pictures held out
optimiser	plain Adam 2e-4, cosine schedule, 3% warmup, batch 8 x 2, two epochs (2,724 steps, 2.67 s a step, 121 minutes on one RTX PRO 6000), the loss on the answer only, seed 20261030
thinking	off in training and at inference through the template's own switch (`enable_thinking=False`); with its thinking on, the base model spends about 2,400 characters of thought before a 150-character row and runs out of budget on a fifth of them, so the line is taught and run without it
record	`qwen-symbolic-thinking` v0.1.22, `run_queue("gemma")` from `notebooks/gemma.ipynb`; the run's full log is `train_summary.json` beside the weights

Versions

version	where	tag
Ulisse E4B v1	the root of this repository	`ulisse-v1` (the tag `v1` marks the same release before the line was named)

The family

size	base	the LoRA adapter	the full fine-tune
E4B	google/gemma-4-E4B-it	gemma-4-e4b-think-symbolic-image	not trained (the line ships adapters)
12B	google/gemma-4-12B-it	gemma-4-12b-think-symbolic-image	not trained (the line ships adapters)

The Ulisse line is Gemma 4; its name is Dante's voyager who sails out to see (Inferno XXVI), as Lucia is the lady of light and sight. Both sizes are trained under Lucia v1's recipe (the 12B at a quarter of its learning rate, 5e-5, the one change its encoder-free picture pathway needs) and read on the same held-out sets as the Lucia family (Qwen3.5: 9B, 4B, 2B, 0.8B), whose cards start at AbstractPhil/qwen3.5-9b-think-symbolic-image.