Text Generation
PEFT
Safetensors
GGUF
MLX
English
lora
gemma4
agent-memory
memory
structured-output
json
ollama
contradiction-detection
Eval Results (legacy)
conversational
Instructions to use sebs-clude/CludeMem-e4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use sebs-clude/CludeMem-e4b with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("google/gemma-4-e4b-it") model = PeftModel.from_pretrained(base_model, "sebs-clude/CludeMem-e4b") - MLX
How to use sebs-clude/CludeMem-e4b with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("sebs-clude/CludeMem-e4b") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use sebs-clude/CludeMem-e4b with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf sebs-clude/CludeMem-e4b:Q4_K_M # Run inference directly in the terminal: llama cli -hf sebs-clude/CludeMem-e4b:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf sebs-clude/CludeMem-e4b:Q4_K_M # Run inference directly in the terminal: llama cli -hf sebs-clude/CludeMem-e4b:Q4_K_M
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf sebs-clude/CludeMem-e4b:Q4_K_M # Run inference directly in the terminal: ./llama-cli -hf sebs-clude/CludeMem-e4b:Q4_K_M
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf sebs-clude/CludeMem-e4b:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama-cli -hf sebs-clude/CludeMem-e4b:Q4_K_M
Use Docker
docker model run hf.co/sebs-clude/CludeMem-e4b:Q4_K_M
- LM Studio
- Jan
- vLLM
How to use sebs-clude/CludeMem-e4b with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "sebs-clude/CludeMem-e4b" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "sebs-clude/CludeMem-e4b", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/sebs-clude/CludeMem-e4b:Q4_K_M
- Ollama
How to use sebs-clude/CludeMem-e4b with Ollama:
ollama run hf.co/sebs-clude/CludeMem-e4b:Q4_K_M
- Unsloth Desktop
- Pi
How to use sebs-clude/CludeMem-e4b with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "sebs-clude/CludeMem-e4b"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "sebs-clude/CludeMem-e4b" } ] } } }Run Pi
# Start Pi in your project directory: pi
- MLX LM
How to use sebs-clude/CludeMem-e4b with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "sebs-clude/CludeMem-e4b"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "sebs-clude/CludeMem-e4b" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "sebs-clude/CludeMem-e4b", "messages": [ {"role": "user", "content": "Hello"} ] }' - Docker Model Runner
How to use sebs-clude/CludeMem-e4b with Docker Model Runner:
docker model run hf.co/sebs-clude/CludeMem-e4b:Q4_K_M
- Lemonade
How to use sebs-clude/CludeMem-e4b with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull sebs-clude/CludeMem-e4b:Q4_K_M
Run and chat with the model
lemonade run user.CludeMem-e4b-Q4_K_M
List all available models
lemonade list
- Hermes Agent
How to use sebs-clude/CludeMem-e4b with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "sebs-clude/CludeMem-e4b"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default sebs-clude/CludeMem-e4b
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use sebs-clude/CludeMem-e4b with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "sebs-clude/CludeMem-e4b"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "sebs-clude/CludeMem-e4b" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Download prompts.json from sebs-clude/CludeMem-e4b: direct link, hf CLI and curl.
- Browser
- Download file 15.5 kB
-
https://huggingface.co/sebs-clude/CludeMem-e4b/resolve/main/prompts.json
- Command line
-
hf download hf://sebs-clude/CludeMem-e4b/prompts.json
-
curl -L -o prompts.json https://huggingface.co/sebs-clude/CludeMem-e4b/resolve/main/prompts.json
15.5 kB
| { | |
| "model": "CludeMem-E4B", | |
| "about": "One entry per memory operation. Send the task's system_prompt as the system message and the input, formatted as user_format describes, as the user message; the reply is one JSON object with output_keys. output_values lists the values those fields take in the training targets. The examples are test items whose inputs do not occur in the training data; example.output is the reference reply.", | |
| "chat_template": { | |
| "apply_chat_template_kwargs": { | |
| "add_generation_prompt": true, | |
| "enable_thinking": true | |
| }, | |
| "rendered_form": "<bos><|turn>system\n<|think|>\n{system}<turn|>\n<|turn>user\n{user}<turn|>\n<|turn>model\n", | |
| "note": "enable_thinking=True reproduces the training prompt: a '<|think|>' line opens the system turn. The model does not emit a thinking block; it replies with one JSON object. transformers' apply_chat_template omits the line unless enable_thinking=True is passed.", | |
| "tokenizers_checked": [ | |
| "google/gemma-4-e4b-it", | |
| "unsloth/gemma-4-E4B-it" | |
| ] | |
| }, | |
| "decoding": "greedy (temperature 0)", | |
| "tasks": { | |
| "CLASSIFY": { | |
| "system_prompt": "Classify the memory. Output JSON: {type, importance (0-1), tags[], concepts[], emotional_valence (-1..1)}.", | |
| "user_format": { | |
| "template": "{memory text}", | |
| "description": "One memory statement as free text on a single line." | |
| }, | |
| "output_keys": [ | |
| "type", | |
| "importance", | |
| "tags", | |
| "concepts", | |
| "emotional_valence" | |
| ], | |
| "output_values": { | |
| "type": [ | |
| "episodic", | |
| "semantic" | |
| ], | |
| "importance": [ | |
| 0.3, | |
| 0.4, | |
| 0.8, | |
| 0.6, | |
| 0.5, | |
| 0.85, | |
| 0.9 | |
| ], | |
| "emotional_valence": [ | |
| 0.2, | |
| 0, | |
| -0.3, | |
| 0.3, | |
| 0.5, | |
| -0.4 | |
| ] | |
| }, | |
| "example": { | |
| "system": "Classify the memory. Output JSON: {type, importance (0-1), tags[], concepts[], emotional_valence (-1..1)}.", | |
| "user": "By the way, around then on 2026-02-16, Priya spent the afternoon reviewing the Polaris roadmap, triaged the mobile backlog, and noted follow-ups for the data sync. It came up while we were chatting.", | |
| "output": { | |
| "type": "episodic", | |
| "importance": 0.3, | |
| "tags": [ | |
| "event", | |
| "task", | |
| "priya" | |
| ], | |
| "concepts": [ | |
| "event", | |
| "task" | |
| ], | |
| "emotional_valence": 0.2 | |
| } | |
| } | |
| }, | |
| "EXTRACT": { | |
| "system_prompt": "Extract atomic memories from the text. Output JSON: {memories:[{content, summary, type}]}.", | |
| "user_format": { | |
| "template": "{Speaker}: {utterance}\\n{Speaker}: {utterance}\\n...", | |
| "description": "A dialogue block, one 'Name: utterance' line per turn. Some rows mix in non-dialogue lines (log lines, markdown tables, tool-call JSON, code fences) that carry no memory." | |
| }, | |
| "output_keys": [ | |
| "memories" | |
| ], | |
| "output_values": { | |
| "memories[].type": [ | |
| "episodic", | |
| "semantic" | |
| ] | |
| }, | |
| "example": { | |
| "system": "Extract atomic memories from the text. Output JSON: {memories:[{content, summary, type}]}.", | |
| "user": "Tomas: Actually, scratch what I said before — tomas now prefers matcha, having switched from cola.\nTomas: By the way, around then tomas started a new job at Brightwell on 2026-07-18. It came up while we were chatting.\n{\"tool\":\"search\",\"status\":\"ok\",\"hits\":1}", | |
| "output": { | |
| "memories": [ | |
| { | |
| "content": "Tomas now prefers matcha, having switched from cola.", | |
| "summary": "Tomas now prefers matcha, having switched from cola.", | |
| "type": "semantic" | |
| }, | |
| { | |
| "content": "Tomas started a new job at Brightwell on 2026-07-18.", | |
| "summary": "Tomas started a new job at Brightwell on 2026-07-18.", | |
| "type": "episodic" | |
| } | |
| ] | |
| } | |
| } | |
| }, | |
| "ENTITIES": { | |
| "system_prompt": "Extract entities and relations. Output JSON: {entities:[{name,type,aliases}], relations:[{head,type,tail}]}.", | |
| "user_format": { | |
| "template": "{memory text}", | |
| "description": "One memory statement as free text on a single line." | |
| }, | |
| "output_keys": [ | |
| "entities", | |
| "relations" | |
| ], | |
| "output_values": { | |
| "entities[].type": [ | |
| "person", | |
| "location", | |
| "organization", | |
| "project" | |
| ], | |
| "relations[].type": [ | |
| "lives_in", | |
| "works_at", | |
| "leads" | |
| ] | |
| }, | |
| "example": { | |
| "system": "Extract entities and relations. Output JSON: {entities:[{name,type,aliases}], relations:[{head,type,tail}]}.", | |
| "user": "Actually, scratch what I said before — on 2026-05-02, Theo spent the afternoon reviewing the Beacon roadmap, triaged the mobile backlog, and noted follow-ups for the product sync.", | |
| "output": { | |
| "entities": [ | |
| { | |
| "name": "Theo", | |
| "type": "person", | |
| "aliases": [] | |
| } | |
| ], | |
| "relations": [] | |
| } | |
| } | |
| }, | |
| "TEMPORAL": { | |
| "system_prompt": "Extract the event date and its temporal links to the known events. Output JSON: {event_date, precision, links:[{type,target}]}.", | |
| "user_format": { | |
| "template": "Today is {YYYY-MM-DD}.\\nNew event [fN]: {text}\\nKnown events:\\n[fN] {text} ({YYYY-MM-DD})\\n...", | |
| "description": "Reference date, the new event with its id, then the known events, each followed by its date in parentheses. A few rows have no known events (the 'Known events:' block is absent)." | |
| }, | |
| "output_keys": [ | |
| "event_date", | |
| "precision", | |
| "links" | |
| ], | |
| "output_values": { | |
| "precision": [ | |
| "day" | |
| ], | |
| "links[].type": [ | |
| "happens_before", | |
| "happens_after" | |
| ] | |
| }, | |
| "example": { | |
| "system": "Extract the event date and its temporal links to the known events. Output JSON: {event_date, precision, links:[{type,target}]}.", | |
| "user": "Today is 2026-06-01.\nNew event [f57]: I think on 2026-03-19, Nina spent the afternoon reviewing the Quartz roadmap, triaged the mobile backlog, and noted follow-ups for the data sync. At least that's my sense of it.\nKnown events:\n[f25] On 2026-03-01, Nina spent the afternoon reviewing the Beacon roadmap, triaged the infra backlog, and noted follow-ups for the data sync. (2026-03-01)\n[f41] On 2026-03-01, Nina spent the afternoon reviewing the Quartz roadmap, triaged the growth backlog, and noted follow-ups for the data sync. (2026-03-01)\n[f23] On 2026-03-10, Nina spent the afternoon reviewing the Sable roadmap, triaged the growth backlog, and noted follow-ups for the design sync. (2026-03-10)\n[f22] On 2026-03-24, Nina spent the afternoon reviewing the Quartz roadmap, triaged the mobile backlog, and noted follow-ups for the platform sync. (2026-03-24)\n[f14] On 2026-04-04, Nina spent the afternoon reviewing the Beacon roadmap, triaged the mobile backlog, and noted follow-ups for the ops sync. (2026-04-04)\n[f34] On 2026-04-06, Nina spent the afternoon reviewing the Echo roadmap, triaged the infra backlog, and noted follow-ups for the design sync. (2026-04-06)", | |
| "output": { | |
| "event_date": "2026-03-19", | |
| "precision": "day", | |
| "links": [ | |
| { | |
| "type": "happens_after", | |
| "target": "f25" | |
| }, | |
| { | |
| "type": "happens_after", | |
| "target": "f41" | |
| }, | |
| { | |
| "type": "happens_after", | |
| "target": "f23" | |
| }, | |
| { | |
| "type": "happens_before", | |
| "target": "f22" | |
| }, | |
| { | |
| "type": "happens_before", | |
| "target": "f14" | |
| }, | |
| { | |
| "type": "happens_before", | |
| "target": "f34" | |
| } | |
| ] | |
| } | |
| } | |
| }, | |
| "CONSOLIDATE": { | |
| "system_prompt": "Consolidate the memories into evidence-linked insights. Output JSON: {insights:[{content, evidence[]}]}.", | |
| "user_format": { | |
| "template": "[fN] {memory}\\n[fN] {memory}\\n...", | |
| "description": "One memory per line, each prefixed with its id." | |
| }, | |
| "output_keys": [ | |
| "insights" | |
| ], | |
| "output_values": {}, | |
| "example": { | |
| "system": "Consolidate the memories into evidence-linked insights. Output JSON: {insights:[{content, evidence[]}]}.", | |
| "user": "[f1] Kai lives in Austin.\n[f2] Kai moved to Seoul on 2026-10-21.\n[f3] Kai currently resides in Seoul.\n[f4] Kai enjoys living in Seoul.\n[f5] Kai prefers cola over matcha.\n[f6] Kai now prefers matcha, having switched from cola.\n[f7] Kai started a new job at Aperture on 2026-08-16.\n[f8] Kai is allergic to gluten.", | |
| "output": { | |
| "insights": [ | |
| { | |
| "content": "Kai lives in Austin.", | |
| "evidence": [ | |
| "f1" | |
| ] | |
| }, | |
| { | |
| "content": "Kai currently resides in Seoul.", | |
| "evidence": [ | |
| "f3" | |
| ] | |
| }, | |
| { | |
| "content": "Kai enjoys living in Seoul.", | |
| "evidence": [ | |
| "f4" | |
| ] | |
| }, | |
| { | |
| "content": "Kai prefers cola over matcha.", | |
| "evidence": [ | |
| "f5" | |
| ] | |
| }, | |
| { | |
| "content": "Kai now prefers matcha, having switched from cola.", | |
| "evidence": [ | |
| "f6" | |
| ] | |
| }, | |
| { | |
| "content": "Kai is allergic to gluten.", | |
| "evidence": [ | |
| "f8" | |
| ] | |
| } | |
| ] | |
| } | |
| } | |
| }, | |
| "COMPACT": { | |
| "system_prompt": "Compact the old memories into one summary, preserving entities and the date range. Output JSON: {summary, preserved_entities[], date_range:{start,end}}.", | |
| "user_format": { | |
| "template": "[fN] {memory}\\n[fN] {memory}\\n...", | |
| "description": "One old memory per line, each prefixed with its id." | |
| }, | |
| "output_keys": [ | |
| "summary", | |
| "preserved_entities", | |
| "date_range" | |
| ], | |
| "output_values": {}, | |
| "example": { | |
| "system": "Compact the old memories into one summary, preserving entities and the date range. Output JSON: {summary, preserved_entities[], date_range:{start,end}}.", | |
| "user": "[f2] Aisha moved to Lisbon on 2026-03-18.\n[f7] Aisha started a new job at Northstar on 2026-10-22.\n[f9] The Beacon project shipped on 2026-04-13.", | |
| "output": { | |
| "summary": "Aisha: 3 events covering Aisha, Lisbon, Northstar, Beacon.", | |
| "preserved_entities": [ | |
| "Aisha", | |
| "Lisbon", | |
| "Northstar", | |
| "Beacon" | |
| ], | |
| "date_range": { | |
| "start": "2026-03-18", | |
| "end": "2026-10-22" | |
| } | |
| } | |
| } | |
| }, | |
| "RECONCILE": { | |
| "system_prompt": "Decide how the two memories relate. Output JSON: {verdict, resolution, weaker_id, confidence}.", | |
| "user_format": { | |
| "template": "Memory A [fN]: {text}\\nMemory B [fN]: {text}", | |
| "description": "Exactly two lines, the two memories to compare with their ids. weaker_id in the output names one of these ids, or is null." | |
| }, | |
| "output_keys": [ | |
| "verdict", | |
| "resolution", | |
| "weaker_id", | |
| "confidence" | |
| ], | |
| "output_values": { | |
| "verdict": [ | |
| "contradicts", | |
| "supersedes", | |
| "duplicate", | |
| "consistent" | |
| ], | |
| "confidence": [ | |
| 0.9, | |
| 0.8, | |
| 0.75, | |
| 0.95, | |
| 0.85 | |
| ], | |
| "resolution": [ | |
| "Memory A and Memory B cannot be ordered; the conflict is unresolved.", | |
| "Memory B conflicts with Memory A; treat B as current.", | |
| "Memory B is the current truth; Memory A is outdated.", | |
| "Memory B restates Memory A; keep one.", | |
| "Memory A and Memory B agree; both are kept.", | |
| "Separate periods; both memories are kept.", | |
| "Different attributes; both memories are kept.", | |
| "Memory A is dated later than Memory B; treat Memory A as current.", | |
| "Memory B is dated later than Memory A; treat Memory B as current.", | |
| "Different people; both memories are kept.", | |
| "The memories overlap in time and cannot be ordered; the conflict is unresolved.", | |
| "Memory B restates Memory A with a coarser date; keep Memory A.", | |
| "Memory A is outdated; Memory B is the current truth.", | |
| "Memory A restates Memory B with a coarser date; keep Memory B.", | |
| "Memory B restates Memory A using another form of the same value; keep Memory A.", | |
| "Memory B is outdated; Memory A is the current truth.", | |
| "Memory A restates Memory B using another form of the same value; keep Memory B.", | |
| "Memory B names the same person as Memory A by a short form; keep Memory A.", | |
| "Memory A names the same person as Memory B by a short form; keep Memory B." | |
| ], | |
| "weaker_id": [ | |
| "one of the two input ids", | |
| "null" | |
| ] | |
| }, | |
| "example": { | |
| "system": "Decide how the two memories relate. Output JSON: {verdict, resolution, weaker_id, confidence}.", | |
| "user": "Memory A [f2]: Caleb moved to Toronto on 2026-03-06.\nMemory B [f4]: Caleb enjoys living in Toronto.", | |
| "output": { | |
| "verdict": "consistent", | |
| "resolution": "Memory A and Memory B agree; both are kept.", | |
| "weaker_id": null, | |
| "confidence": 0.9 | |
| } | |
| } | |
| }, | |
| "QUERY": { | |
| "system_prompt": "Understand the query. Output JSON: {expanded_queries[], temporal_constraints, type_filters[], entities[], intent}.", | |
| "user_format": { | |
| "template": "Today is {YYYY-MM-DD}.\\n{question}", | |
| "description": "Reference date on the first line, the user's question on the second." | |
| }, | |
| "output_keys": [ | |
| "expanded_queries", | |
| "temporal_constraints", | |
| "type_filters", | |
| "entities", | |
| "intent" | |
| ], | |
| "output_values": { | |
| "intent": [ | |
| "lookup", | |
| "temporal_lookup" | |
| ], | |
| "type_filters": [ | |
| [] | |
| ] | |
| }, | |
| "example": { | |
| "system": "Understand the query. Output JSON: {expanded_queries[], temporal_constraints, type_filters[], entities[], intent}.", | |
| "user": "Today is 2026-06-01.\nDid Zara move to Lyon on 2026-04-19?", | |
| "output": { | |
| "expanded_queries": [ | |
| "Did Zara move to Lyon on 2026-04-19?", | |
| "Did Zara move to Lyon on 2026-04-19" | |
| ], | |
| "temporal_constraints": { | |
| "after": null, | |
| "before": null | |
| }, | |
| "type_filters": [], | |
| "entities": [ | |
| "Zara", | |
| "Lyon" | |
| ], | |
| "intent": "temporal_lookup" | |
| } | |
| } | |
| }, | |
| "ANSWER": { | |
| "system_prompt": "Answer ONLY from the provided memories; abstain if unsupported. Output JSON: {rationale, answer, citations[], confidence, abstain}.", | |
| "user_format": { | |
| "template": "Question: {question}\\n\\nMemories:\\n[fN] {memory}\\n[fN] {memory}\\n...", | |
| "description": "The question, a blank line, 'Memories:', then one retrieved memory per line with its id. citations in the output are ids from this list." | |
| }, | |
| "output_keys": [ | |
| "rationale", | |
| "answer", | |
| "citations", | |
| "confidence", | |
| "abstain" | |
| ], | |
| "output_values": { | |
| "abstain": [ | |
| false, | |
| true | |
| ], | |
| "confidence": [ | |
| 0.9, | |
| 0.8 | |
| ], | |
| "answer (when abstain is true)": [ | |
| "I do not have enough information to answer that." | |
| ] | |
| }, | |
| "example": { | |
| "system": "Answer ONLY from the provided memories; abstain if unsupported. Output JSON: {rationale, answer, citations[], confidence, abstain}.", | |
| "user": "Question: Did Tomas move to Lisbon on 2026-03-14?\n\nMemories:\n[f1] Tomas lives in Toronto.\n[f2] Tomas moved to Lisbon on 2026-03-12.", | |
| "output": { | |
| "rationale": "No provided memory answers this.", | |
| "answer": "I do not have enough information to answer that.", | |
| "citations": [], | |
| "confidence": 0.8, | |
| "abstain": true | |
| } | |
| }, | |
| "example_answerable": { | |
| "system": "Answer ONLY from the provided memories; abstain if unsupported. Output JSON: {rationale, answer, citations[], confidence, abstain}.", | |
| "user": "Question: Where does Maya live now?\n\nMemories:\n[f2] Maya moved to Austin on 2026-10-22.\n[f1] Maya lives in Lisbon.\n[f3] Maya currently resides in Austin.", | |
| "output": { | |
| "rationale": "Supported by f2.", | |
| "answer": "Austin", | |
| "citations": [ | |
| "f2" | |
| ], | |
| "confidence": 0.9, | |
| "abstain": false | |
| } | |
| } | |
| } | |
| } | |
| } | |