system stringclasses 6
values | instruction stringlengths 383 55.4k | output stringlengths 109 12.5k | metadata unknown |
|---|---|---|---|
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
25yo software engineer & senior UX designer in Milan, Italy. Fluent Italian. Values family, authenticity, independence, privacy rights. Recently lost grandmother, Rachel's best friend's mom, John's dad. Close with best friend Rachel, cousin Rachel, sister. Strained relationship with controlling father (bus... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for a specific commitment Patagonia made in 2019 regarding tax savings from a conversation about balancing stakeholder needs. The Core memory only says the user admires Patagonia and values stakeholder balance, transparency, and employee empowerme... | {
"row_id": "v5:8:run18-hybrid-evidence-drop:gpt4_1d80365e__64a6e2bdf1:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 684,
"source": "callback-decision",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidenc... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Marketing Specialist at XYZ Corporation (2+ years), reports to Rachel. Python, Pandas, Matplotlib, Seaborn. Planning data analytics certification.
Construction lawyer (NSW): building regulations, contracts, homeowner protection, educational content.
Full-stack developer: React, JavaScript, Node.js, Mongo... | <tool_call>
{"name":"search_memory","arguments":{"query":"automated formula prompts UI process product characteristics audience selection","top_k":8,"banks":["episodic","semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:v3:train:6ade9755:synthetic:membuilder_flat_after_v1:sft:6ade9755:S35:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:6ade9755:... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User Profile:
- English instructor at Texas community college (Distance Ed/Educational Technology); detail-oriented, methodical, data-driven
- Bilingual English/Spanish; learning Mandarin
- Lives on Indiana farm; moved to UK for Edinburgh study abroad (visa approved May 27, 2023)
Family: Brother graduated... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The retrieved entries dated 27 May 2023 directly address applicants from India: m000251 and m000243 say Japanese Encephalitis vaccine may be recommended if the person is from rural India or spent significant time in rural areas, Hepatitis B vaccination should be up... | {
"row_id": "v5:6:run29ds-stale-core:d01c6aa8__96f8754925:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 6,
"run": "run29ds-stale-core",
"trajectory": "runs/run29ds-stale-core/d01c6aa8__96f8754925.json... |
Manage MemBuilder episodic memory.
You are the Episodic Memory Manager. Manage time-ordered event memories.
Episodic Memory stores time-ordered, event-based information from interactions—essentially, the "diary" of user events.
Each episodic memory MUST include:
(a) summary: Short textual summary of the event (concis... | {"bank": "episodic", "conversation_timestamp": "2023/05/21 (Sun) 11:03", "messages": "[REAL_USER]: I'm looking for some advice on interior design. I just rearranged my living room furniture and added a new coffee table, but I'm not sure what kind of decorations would go well with it.\n[AI_ASSISTANT]: Exciting! Rearrang... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"episodic","operations":[{"action":"UPDATE","old_memory":"[EPISODIC] 2023-05-19: Real User organized closet after months of procrastination | Details: Last Sunday from conversation date of 2023-05-21 11:03 (calculated as 2023-05-19), Real User finally organized th... | {
"row_id": "v5:2:run29ds-value-drop:4bc144e2__d9e2e90a45:2:1",
"row_role": "episodic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 2,
"run": "run29ds-value-drop",
"trajectory": "runs/run29ds-value-drop/4bc144e2__d9e2e90a45.js... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Tzu-li Liu. Neuroscience/biomedical engineering researcher (brain-machine interfaces, neuromodulation). Software sales using MEDDPICC. Writing thesis on social media marketing impact. Uses Obsidian, detail-oriented, data-driven.
Creative: Sculptor (clay/metal/wood), guitarist learning violin (Suzuki metho... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000130","m000452","m000129","m000131","m000132","m000133"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:9:run29ds-distractor-rank:gpt4_15e38248__667f562f4c:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 9,
"run": "run29ds-distractor-rank",
"trajectory": "runs/run29ds-distractor-rank/gpt4... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Creative professional near Denver. Married, family with countryside lake house.
Interests: AI image generation (dystopian/anime), BJJ white belt, blockchain/crypto (Flow), game dev with NFT/Web3. Philosophy: Alan Watts, Hindu mythology (Kali), Buddhism. Creates spirituality manuscripts/YouTube scripts.
M... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000554","m000553","m000372"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:2:run29ds-value-drop:1c0ddc50__99103baaa3:3:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 2,
"run": "run29ds-value-drop",
"trajectory": "runs/run29ds-value-drop/1c0ddc50__99103baaa3.j... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
**CRITICAL: This is an asylum seeker in Canada, NOT Stephen Donohoe**
Asylum seeker in Canada. Received asylum approval February 20, 2023. Location in Canada unspecified.
Language Learning: Attending LINC program since November 2020, currently level 5. Teacher: Mrs. Patel. Made friends including someone ... | <tool_call>
{"name":"search_memory","arguments":{"query":"bathroom deep cleaning order steps Sunday","top_k":5,"banks":["episodic","procedural"]}}
</tool_call> | {
"row_id": "v5:v3:train:06db6396:synthetic:membuilder_flat_after_v1:sft:06db6396:S49:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:06db6396:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Emily Rose Thompson (post-divorce). Founder DesignSpark (Jan 15), creates Brand Style Guide Templates. Builds remote teams prioritizing emotional intelligence. Thesis: social media's mental health impact.
Projects: "Black Community Impact" gamification, Win'it e-commerce. Designs APIs/databases. Developin... | <tool_call>
{"name":"search_memory","arguments":{"query":"motorcycle upgrades exhaust suspension tuning bike","top_k":10,"banks":["episodic","semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:15:run29ds-unanswerable:1e043500__f0e65f49c2:3:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 15,
"run": "run29ds-unanswerable",
"trajectory": "runs/run29ds-unanswerable/1e043500__f0e6... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Bilingual (English/Turkish). Developing AI fashion tech: Fashionoscope (B2B) and Stiloni (B2C mobile app). Product management/MVP expertise. Background: NSW residential construction law/consulting, HIA contracts, marketing/copywriting.
New Haven, CT homeowner (Jan 2023). Vegan, budget-conscious, environme... | <tool_call>
{"name":"search_memory","arguments":{"query":"Convict Lake Campground elevation","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_fa19884d:synthetic:membuilder_flat_after_v1:sft:gpt4_fa19884d:S37:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
Manage MemBuilder episodic memory.
You are the Episodic Memory Manager. Manage time-ordered event memories.
Episodic Memory stores time-ordered, event-based information from interactions—essentially, the "diary" of user events.
Each episodic memory MUST include:
(a) summary: Short textual summary of the event (concis... | {"bank": "episodic", "conversation_timestamp": "2023/05/28 (Sun) 04:11", "messages": "[REAL_USER]: I'm planning a summer trip and considering Costa Rica or Iceland. Can you give me some tips on what to do and see in these countries? By the way, I just got back from a trip to Japan and before that, I had a blast on a 4-... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"episodic","operations":[{"action":"ADD","memory":"2023-05: User returned from a trip to Japan | Details: Just before conversation on 2023-05-28 (calculated as 2023-05), user returned from a trip to Japan."},{"action":"ADD","memory":"2023-01: User went on a 4-day ... | {
"row_id": "v5:4:run26-memory-review-checklist:3ba21379__23__episodic:negative",
"row_role": "episodic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "hard-negative",
"error_type": 4,
"run": "run26-memory-review-checklist",
"trajectory": "runs/run26-memory-review-checklist/3ba21379_... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000276","m000275","m000277","m000253","m000249"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:91b15a6e:synthetic:membuilder_flat_after_v1:sft:91b15a6e:S29:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:91b15a6e:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
**Personal:** Social, health-conscious calorie tracker. Friend: John. Cherokee heritage. Zip 12345 (California).
**Professional:** Marketing expert (campaigns, content, social media, analytics). Tax law specialist Sub-Saharan Africa. Uses APA, Trello, RescueTime.
**Faith:** Choir St. Mary's Church, Bible... | <tool_call>
{"name":"search_memory","arguments":{"query":"reggae artists collaborated brands apparel clothing lines","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:73d42213:synthetic:membuilder_flat_after_v1:sft:73d42213:S23:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:73d42213:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Stanford CS Master's (AI/healthcare, 5 papers, 100+ citations). Hebrew University student (ID: 11822678). Seeking AI engineering at Google/Microsoft. Skills: Python, TensorFlow, PyTorch, CV, NLP, AWS. Research: Medical imaging. Freelance AI/ML tech writing interest. Works in Cloud Economics/cloud computing... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000011","m000005","m000010","m000018"],"revisit_agent":"memory","feedback":"Three searches (inherited vintage Rolex year made; grandfather Rolex provenance/serial number; Rolex manufacture date/model/appraisal) all return only 'mid-20th-century vintage Rolex' wi... | {
"row_id": "v5:2:run29ds-value-drop:3ba21379__88115ed8f3:1:3",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 97,
"source": "callback-decision",
"error_type": 2,
"run": "run29ds-value-drop",
"trajectory": "runs/run29ds-value-drop/3ba21379__88115ed8f3... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User Profile:
**Professional:**
- Runs Renewable Leads (lead gen) and e-commerce (sustainable home goods)
- Location: 1 Askew Cottages, Beaver Place, Worksop, S80 2ER
- Email: datacontroller@renewablelife.com
- Focus: GDPR compliance, circular economy, sustainability
- Works from home, ~10 colleagues
- Pr... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000363","m000362","m000304","m000305","m000306","m000446","m000250","m000239","m000300","m000048"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:7:run29ds-retrieval-gap:dfde3500__901e735f77:2:3",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 7,
"run": "run29ds-retrieval-gap",
"trajectory": "runs/run29ds-retrieval-gap/dfde3500__901... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/28 (Sun) 13:20", "messages": "[REAL_USER]: How many Missouri Valley Conference teams made it to the Sweet Sixteen in the past five years?\n[AI_ASSISTANT]: Three Missouri Valley Conference teams have made it to the Sweet Sixteen in the past five years. These teams incl... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"REPLACE","old_text":"Sports: College basketball fan, March Madness. Follows Missouri Valley Conference, especially Loyola Chicago's 2018 Final Four.","new_text":"Sports: College basketball fan, March Madness. Follows Missouri Val... | {
"row_id": "v5:1:run18-hybrid-memory-gap:4bc144e2__660b9b18a9:2:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 1,
"run": "run18-hybrid-memory-gap",
"trajectory": "runs/run18-hybrid-memory-gap/4bc144e2__66... |
Manage MemBuilder episodic memory.
You are the Episodic Memory Manager. Manage time-ordered event memories.
Episodic Memory stores time-ordered, event-based information from interactions—essentially, the "diary" of user events.
Each episodic memory MUST include:
(a) summary: Short textual summary of the event (concis... | {"bank": "episodic", "conversation_timestamp": "2023/05/27 (Sat) 17:13", "messages": "[REAL_USER]: I'm looking for some inspiration for my next photography project. I recently attended a workshop on capturing cityscapes at night, but I'm interested in exploring other themes. Do you have any suggestions? By the way, I'v... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"episodic","operations":[{"action":"ADD","memory":"2023-05-27 17:13: User is seeking inspiration for their next photography project | Details: During conversation on 2023-05-27 17:13, user said they are looking for inspiration for their next photography project. T... | {
"row_id": "v5:4:run26-memory-review-checklist:3fdac837__30__episodic:negative",
"row_role": "episodic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "hard-negative",
"error_type": 4,
"run": "run26-memory-review-checklist",
"trajectory": "runs/run26-memory-review-checklist/3fdac837_... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/30 (Tue) 18:15", "messages": "[REAL_USER]: Hello friend. My task for today is to engage in a debate with you. Will you humor me in this regard?\n[AI_ASSISTANT]: Of course! I'd be happy to engage in a debate with you. What topic would you like to debate about?\n[REAL_U... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Philosophy & Debate: Enjoys intellectual debates and philosophical discussions. Familiar with critical theory, particularly Theodor Adorno's work and concept of \"negative dialectics\". Interested in theoria (t... | {
"row_id": "v5:v3:train:b6025781:46:core:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:b6025781:46:core:0",
"prompt_tokens": 2598,
"target_tokens": 114,
"trained_tokens": 114,
"total... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Bilingual (English/Turkish). Developing AI fashion tech: Fashionoscope (B2B) and Stiloni (B2C mobile app). Product management/MVP expertise. Background: NSW residential construction law/consulting, HIA contracts, marketing/copywriting.
New Haven, CT homeowner (Jan 2023). Vegan, budget-conscious, environme... | <tool_call>
{"name":"search_memory","arguments":{"query":"Swedish dishcloths cotton cellulose","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_fa19884d:synthetic:membuilder_flat_after_v1:sft:gpt4_fa19884d:S5:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Name: Alister. Chinese speaker in Tokyo (5+ months, May 2023). Geophysicist using Python/pandas, Kubernetes. Researches anti-microbial peptides/proteins, ML, molecular dynamics, structural bioinformatics.
Personality: Detail-oriented, methodical, budget-conscious, values community service. Not tech-savvy,... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000069","m000065","m000074","m000063","m000070"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:3fdac837:synthetic:membuilder_flat_after_v1:sft:3fdac837:S6:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:3fdac837:s... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User Profile:
Name: Aman Ayaz
Contact: 647-717-5780, aman.ayaz44@gmail.com
Location: Tokyo, Japan (Roppongi, since Nov 2022)
Education & Career:
- 3rd-year Electrical Engineering, Toronto Metropolitan University
- MBA U Michigan (2021), Harvard CS50 (2021)
- Toronto Met Formula Racing (electrical drivetr... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The evidence shows an education trust fund for the child (m000093 and related entries) and, in Core memory, graduate school in CS at Stanford/CMU/MIT/Berkeley/UW as a higher-education consideration.","answer":"An education trust fund and graduate school (CS at Stan... | {
"row_id": "v5:v3:train:0862e8bf:synthetic:membuilder_flat_after_v1:sft:0862e8bf:S13:72df23d381d94d99a65da6fef7c86bbb:4:answer:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:0862e8bf:synthe... |
Manage MemBuilder semantic memory.
You are the Semantic Memory Manager. Manage conceptual knowledge about people, places, objects, and concepts.
Semantic Memory holds general knowledge, concepts, definitions, and facts. It is the storehouse of abstract understanding about the world.
IMPORTANT: ONLY save NEW concepts ... | {"bank": "semantic", "conversation_timestamp": "2023/11/08 (Wed) 18:57", "messages": "[REAL_USER]: How has globalization affected the lives and livelihoods of people in different parts of the world?\n[AI_ASSISTANT]: Globalization has had a profound impact on people's lives and livelihoods in different parts of the worl... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"semantic","operations":[{"action":"ADD","memory":"Globalization - Five main impacts on people's lives and livelihoods: economic, cultural, environmental, political, and technological changes | Details: Discussed on 2023-11-08 as the answer to how globalization ha... | {
"row_id": "v5:10:run36-distractor-core:06db6396__027ae7b4f6:2:3",
"row_role": "semantic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 10,
"run": "run36-distractor-core",
"trajectory": "runs/run36-distractor-core/06db6396__02... |
Manage MemBuilder procedural memory.
You are the Procedural Memory Manager. Manage step-by-step processes, workflows, and instructions.
Procedural Memory contains how-to guides, step-by-step instructions, or processes the user might follow.
Each procedural memory entry MUST include:
(a) entry_type: Type of procedure ... | {"bank": "procedural", "conversation_timestamp": "2023/05/23 (Tue) 16:41", "messages": "[REAL_USER]: what are some lesser known advice from popular therapists\n[AI_ASSISTANT]: 1. \"The best thing you can do for the people you love is to take care of yourself.\" - Brené Brown\n2. \"The opposite of addiction is not sobri... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"procedural","operations":[{"action":"ADD","memory":"How to build a news summarization startup similar to Inshorts | Steps: 1. Research and identify the target audience - understand the user base and the kind of news they are interested in, identify the target mar... | {
"row_id": "v5:v3:train:gpt4_b4a80587:31:procedural:0",
"row_role": "procedural",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:gpt4_b4a80587:31:procedural:0",
"prompt_tokens": 5604,
"target_tokens": 303,
"tr... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Photographer: Canon EOS 80D, 50mm f/1.8. Shoots portraits/landscapes. Uses Luminar, Capture One.
Yoga/Wellness: 6mo practice (Vinyasa, Yin). Studies pranayama, Sutras, Bhagavad Gita. Meditates for stress.
Food: Low spice tolerance. Loves avocado toast, NYC brunch. Vegetarian-interested, sustainable agric... | <tool_call>
{"name":"search_memory","arguments":{"query":"government corruption prevention top leadership accountability","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:8752c811:synthetic:membuilder_flat_after_v1:sft:8752c811:S14:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:8752c811:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
QA Automation Test Engineer at Wahoo Fitness (remote, Minneapolis, Oct 2021, promoted Nov 2022). Specializes in mobile apps, automation, embedded devices. Skills: Python, pytest, BDD, CI/CD, Jenkins, REST API. Leads bike computer automation. Interested in Sr. QA role at 1Password.
Genealogy researcher: 19... | <tool_call>
{"name":"search_memory","arguments":{"query":"helping a friend find a local art class for their creative development community center art school","top_k":5,"banks":["episodic"]}}
</tool_call> | {
"row_id": "v5:8:run18-hybrid-evidence-drop:945e3d21__c4dc2333a7:3:3",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-dro... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/21 (Sun) 21:05", "messages": "[AI_ASSISTANT]: Another example of how research questions can guide the product development process is in the case of developing a new social media platform. The team might have the following research questions:\n\nWhat features do users ... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"**Learning & Development:** Studying product development and user research methodologies. Interested in defining research goals, conducting user interviews, and creating actionable insights from user research. ... | {
"row_id": "v5:v3:train:73d42213:11:core:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:73d42213:11:core:0",
"prompt_tokens": 3948,
"target_tokens": 147,
"trained_tokens": 147,
"total... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The cycling-plan entries (m000085, m000086, m000098) state that in addition to Saturday morning 20km rides, the user plans to add one or two shorter weekday rides focused on high-intensity interval training (HIIT). The strength-training entries (m000084, m000089) a... | {
"row_id": "v5:14:run30ds-multi-session:91b15a6e__2e5b0135ad:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 544,
"source": "callback-decision",
"error_type": 14,
"run": "run30ds-multi-session",
"trajectory": "runs/run30ds-multi-session/91b15a6e__2... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Software engineer, 5+ years experience. Former Research Officer at ADD Korea (UAV software), Full-Stack Developer at Conalog (cloud-native SaaS, IoT). Team leader applying for Palantir Forward Deployed Software Engineer role. Values diverse experiences, complex problem-solving, societal impact, teamwork, r... | <tool_call>
{"name":"search_memory","arguments":{"query":"United Methodist Church UMC issue congregations impact","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_8279ba03:synthetic:membuilder_flat_after_v1:sft:gpt4_8279ba03:S18:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Senior Software Engineer at TechCorp (agricultural tech), aspiring Technical Lead/Architect. Expertise: data science, ML, Python, AI, cloud computing, Docker, WebSockets, WebRTC, Node.js, React. Building video chat app. Owns CreativeSpark Marketing agency, YouTube/Instagram creator launching Patreon "Marke... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Entry m000190 records the user asking about the most important VFX workflow concepts on 24 May 2023 at 11:14. Entry m000216 records the user asking about machine learning automation for restaurant POS systems on 25 May 2023 at 07:00. From 24 May 2023 to 25 May 2023... | {
"row_id": "v5:7:run18-hybrid-retrieval-gap:gpt4_2ba83207__de96755184:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 7,
"run": "run18-hybrid-retrieval-gap",
"trajectory": "runs/run18-hybrid-retrieval-... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Finance/asset management professional: treasury operations, collateral management, derivatives, regulatory compliance. Retail shop owner. Founder/leader of Codestates (talent transformation company).
Account Executive at OwnBackup (SaaS) since Feb 2021. 10+ years sales. President's Club 2022, Top Performe... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for the scholarship program Beyoncé launched in 2017 to encourage young women to pursue higher education. Retrieved entry m000089 directly states that in 2017 Beyoncé launched the Formation Scholarship Program for this purpose, with detail about t... | {
"row_id": "v5:6:run32-stale-core:86b68151__2d827fa733:4:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 6,
"run": "run32-stale-core",
"trajectory": "runs/run32-stale-core/86b68151__2d827fa733.json",
"... |
Manage MemBuilder procedural memory.
You are the Procedural Memory Manager. Manage step-by-step processes, workflows, and instructions.
Procedural Memory contains how-to guides, step-by-step instructions, or processes the user might follow.
Each procedural memory entry MUST include:
(a) entry_type: Type of procedure ... | {"bank": "procedural", "conversation_timestamp": "2023/05/26 (Fri) 19:01", "messages": "[REAL_USER]: What percentage of Tyne and Wear's workforce is employed in manufacturing?\n[AI_ASSISTANT]: As an AI language model, I don't have access to the latest data. However, according to the Office for National Statistics (ONS)... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"procedural","operations":[],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:0862e8bf:38:procedural:0",
"row_role": "procedural",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:0862e8bf:38:procedural:0",
"prompt_tokens": 5733,
"target_tokens": 31,
"trained_token... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000227","m000226","m000225","m000267","m000262"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_b5700ca0:synthetic:membuilder_flat_after_v1:sft:gpt4_b5700ca0:S22:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Location: San Francisco, CA. Tech industry.
Career: Corporate event planner (networking receptions). Exploring renewable energy, live music production, virtual events. Career transition March 2023. Side business: handcrafted candles at farmers markets.
Spirituality: Questioning Christian faith after gran... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":[],"revisit_agent":"memory","feedback":"Four searches with varied phrasings (DV-2024 entries per registration period, lottery registration rules, one entry per person, entry limit immigration) returned only unrelated results (photography, geomancy, gift cards, AAPI... | {
"row_id": "v5:1:run18-hybrid-memory-gap:b6025781__755c7d957d:1:4",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 58,
"source": "callback-decision",
"error_type": 1,
"run": "run18-hybrid-memory-gap",
"trajectory": "runs/run18-hybrid-memory-gap/b6025... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Emily Rose Thompson (post-divorce). Founder DesignSpark (Jan 15), creates Brand Style Guide Templates. Builds remote teams prioritizing emotional intelligence. Thesis: social media's mental health impact.
Projects: "Black Community Impact" gamification, Win'it e-commerce. Designs APIs/databases. Developin... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Core memory states under Upcoming: 'Baby shower next weekend (sundress, tote, boots outfit)'. This matches the flowy dress (fashion notes flowy sundresses), tote bag, and black Zara boots described in the question. No retrieved entries are needed since Core memory ... | {
"row_id": "v5:2:run25-value-drop:1e043500__400f6cab29:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 2,
"run": "run25-value-drop",
"trajectory": "runs/run25-value-drop/1e043500__400f6cab29.json",
"mo... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
LA resident. Fantasy writer (MK/Wukong, dialogue-driven), studies ecocriticism. Web/software developer (React, Python, Django, Rails) on real estate portal with ML models. Provides copywriting/marketing.
Marathon runner (April 15 PB), training for sprint triathlon. Yoga 2x/week, weekend cyclist. Landscape... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Entry m000458 states the Fog Linen Pillow is filled with a blend of down alternative (recycled polyester, hypoallergenic and cruelty-free) and natural kapok fibers. Entry m000453 confirms the same filling: down alternative made from recycled polyester and natural k... | {
"row_id": "v5:8:run18-hybrid-evidence-drop:7e974930__4f28e015e9:4:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-drop/7... |
Manage MemBuilder semantic memory.
You are the Semantic Memory Manager. Manage conceptual knowledge about people, places, objects, and concepts.
Semantic Memory holds general knowledge, concepts, definitions, and facts. It is the storehouse of abstract understanding about the world.
IMPORTANT: ONLY save NEW concepts ... | {"bank": "semantic", "conversation_timestamp": "2023/05/28 (Sun) 14:11", "messages": "[REAL_USER]: I'm planning a trip to the Eastern Sierra in July or August and was wondering if you could recommend some scenic hiking trails in the area. By the way, I just started my solo camping trip to Yosemite National Park today a... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"semantic","operations":[{"action":"ADD","memory":"User - Current location: Solo camping trip at Yosemite National Park starting May 28, 2023 | Details: The user started a solo camping trip to Yosemite National Park on May 28, 2023 (Sunday). They are excited to ex... | {
"row_id": "v5:v3:train:f523d9fe:36:semantic:0",
"row_role": "semantic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:f523d9fe:36:semantic:0",
"prompt_tokens": 8085,
"target_tokens": 1238,
"trained_tokens": ... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/08/10 (Thu) 01:32", "messages": "[REAL_USER]: How often should a dog be vaccinated against parvo and other common canine diseases, and what are the best vaccination schedules?\n[AI_ASSISTANT]: Dogs should be vaccinated against parvo and other common canine diseases as pu... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Owns a female dog that rarely leaves the house. Cost-conscious about dog vaccinations and skeptical of their necessity; questions skipping boosters, relying on natural immunity, or treating illness instead of p... | {
"row_id": "v5:1:run18-hybrid-memory-gap:4b24c848__bafbc7f429:2:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 1,
"run": "run18-hybrid-memory-gap",
"trajectory": "runs/run18-hybrid-memory-gap/4b24c848__ba... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Finance/asset management professional: treasury operations, collateral management, derivatives, regulatory compliance. Retail shop owner. Founder/leader of Codestates (talent transformation company).
Account Executive at OwnBackup (SaaS) since Feb 2021. 10+ years sales. President's Club 2022, Top Performe... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The matching entry states that in a Pyomo model with Binary variables x and y, the function my_constraint_rule returned the linear constraint x + y <= 1. The other entries either use NonNegativeReals, a different rule function, or a different constraint.","answer":... | {
"row_id": "v5:9:run29ds-distractor-rank:86b68151__580755885a:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 333,
"source": "callback-decision",
"error_type": 9,
"run": "run29ds-distractor-rank",
"trajectory": "runs/run29ds-distractor-rank/86b6815... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Emily Rose Thompson. Recently divorced from John Wilson. SEO expert/marketing copywriter for fashion/jewelry brands (Jared Lang, Kinsley Armelle). Works as 'Jo Peninsulas' with environmental orgs on UN SDGs/climate initiatives with Isaac Cohen. Master's in Computer Science (2021), 2 years in AI/ML, interes... | <tool_call>
{"name":"search_memory","arguments":{"query":"dofollow vs nofollow links blog post three main benefits","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:2:run29ds-value-drop:76d63226__3042be9f6b:3:0",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 2,
"run": "run29ds-value-drop",
"trajectory": "runs/run29ds-value-drop/76d63226__3042be9f6b.j... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User Profile:
- English instructor at Texas community college (Distance Ed/Educational Technology); detail-oriented, methodical, data-driven
- Bilingual English/Spanish; learning Mandarin
- Lives on Indiana farm; moved to UK for Edinburgh study abroad (visa approved May 27, 2023)
Family: Brother graduated... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000032","m000025","m000244","m000035","m000238","m000408","m000234","m000086"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:14:run30ds-multi-session:d01c6aa8__ab30c4c8aa:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 14,
"run": "run30ds-multi-session",
"trajectory": "runs/run30ds-multi-session/d01c6aa8__ab3... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
San Francisco resident (May 2023), later South Australia, now Chicago (May 2023+). Pets: Max (dog), Luna (cat). Vegetarian/vegan, mild foods, managing diabetes. Close with 86-year-old grandmother. Formally diagnosed Autistic, registered with NDIS. Severe anxiety, especially at night and uncertain situation... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The retrieved entries from 26 May 2023 state Kandinsky saw artists as social activists and visionaries with a sacred responsibility to use their talents for positive change, as prophets/seers and spiritual guides who convey spiritual and moral messages and guide hu... | {
"row_id": "v5:v3:train:7527f7e2:synthetic:membuilder_flat_after_v1:sft:7527f7e2:S26:72df23d381d94d99a65da6fef7c86bbb:3:answer:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7527f7e2:synthe... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
LA resident. Fantasy writer (MK/Wukong, dialogue-driven), studies ecocriticism. Web/software developer (React, Python, Django, Rails) on real estate portal with ML models. Provides copywriting/marketing.
Marathon runner (April 15 PB), training for sprint triathlon. Yoga 2x/week, weekend cyclist. Landscape... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000076","m000069","m000077","m000068","m000072"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:7e974930:synthetic:membuilder_flat_after_v1:sft:7e974930:S5:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7e974930:s... |
Manage MemBuilder episodic memory.
You are the Episodic Memory Manager. Manage time-ordered event memories.
Episodic Memory stores time-ordered, event-based information from interactions—essentially, the "diary" of user events.
Each episodic memory MUST include:
(a) summary: Short textual summary of the event (concis... | {"bank": "episodic", "conversation_timestamp": "2023/03/03 (Fri) 10:25", "messages": "[REAL_USER]: If they accept my linkedin request to connect, this is the cadence I use:\n\nLinkedIn Message:\n\nHi {{first\\_name}}, \nThanks for accepting my invite to connect. \nAs mentioned, we just completed a 2 year project with S... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"episodic","operations":[{"action":"ADD","memory":"2023-03-03 10:25: User shared LinkedIn and email cadence strategy for connecting with spacetech companies | Details: Mentioned during conversation on 2023-03-03 10:25, user shared their detailed outreach cadence u... | {
"row_id": "v5:v3:train:80ec1f4f_abs:17:episodic:0",
"row_role": "episodic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:80ec1f4f_abs:17:episodic:0",
"prompt_tokens": 6721,
"target_tokens": 1107,
"trained_t... |
Manage MemBuilder semantic memory.
You are the Semantic Memory Manager. Manage conceptual knowledge about people, places, objects, and concepts.
Semantic Memory holds general knowledge, concepts, definitions, and facts. It is the storehouse of abstract understanding about the world.
IMPORTANT: ONLY save NEW concepts ... | {"bank": "semantic", "conversation_timestamp": "2023/07/11 (Tue) 06:52", "messages": "[REAL_USER]: I'm looking for some fashion advice. I recently got a new pair of distressed denim jeans from Zara last Saturday, which was a week ago, and I'm loving the fit and style. I got them for 20% off, which was a great deal. I'm... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"semantic","operations":[{"action":"ADD","memory":"User - Fashion purchase: Zara distressed denim jeans | Details: User bought a new pair of distressed denim jeans from Zara on Saturday around July 1, 2023 (described as 'last Saturday, which was a week ago' relati... | {
"row_id": "v5:1:run18-hybrid-memory-gap:ce6d2d27__1f2666d744:2:2",
"row_role": "semantic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 1,
"run": "run18-hybrid-memory-gap",
"trajectory": "runs/run18-hybrid-memory-gap/ce6d2d27... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000394","m000392","m000393","m000135","m000443"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_b5700ca0:synthetic:membuilder_flat_after_v1:sft:gpt4_b5700ca0:S39:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question matches entries m000015 and m000024 from the same session (20 May 2023 19:41): the user stated they learned about the concept of 'beggarly prayers' in a sermon at St. Mary's Church and it really resonated with them — approaching God as beggars, acknowl... | {
"row_id": "v5:6:run32-stale-core:91b15a6e__f71babbef1:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 6,
"run": "run32-stale-core",
"trajectory": "runs/run32-stale-core/91b15a6e__f71babbef1.json",
"mo... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
50yo Mumbai entrepreneur, CEO Grailzy (watch auction), co-founder Comparative (no-code analytics). NYC-based B2B sales/outreach for 6sense, writes for CRO/VP SaaS audiences. Tech: Spring Cloud, GraphQL, Angular, TensorFlow.js, Docker, Kubernetes.
Marathon runner (Chicago 4:10, trains 4-5x/week, 20-25mi/we... | <tool_call>
{"name":"search_memory","arguments":{"query":"stories balance negative news social media action","top_k":5,"banks":["episodic","semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:v3:train:e25c3b8d:synthetic:membuilder_flat_after_v1:sft:e25c3b8d:S12:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:e25c3b8d:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Finance/asset management professional: treasury operations, collateral management, derivatives, regulatory compliance. Retail shop owner. Founder/leader of Codestates (talent transformation company).
Account Executive at OwnBackup (SaaS) since Feb 2021. 10+ years sales. President's Club 2022, Top Performe... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000170","m000168","m000167","m000169","m000352"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:86b68151:synthetic:membuilder_flat_after_v1:sft:86b68151:S17:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:86b68151:... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/20 (Sat) 18:30", "messages": "[REAL_USER]: I'm thinking of getting a travel rewards credit card to maximize my points. Can you recommend some popular options that would be suitable for frequent travelers like me?\n[AI_ASSISTANT]: There are many excellent travel reward... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Frequent traveler who has taken 7 flights in the last 9 weeks. Takes train regularly to visit sister in Philadelphia (approximately 2-hour ride from hometown). Interested in travel rewards credit cards to maxim... | {
"row_id": "v5:v3:train:76d63226:3:core:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:76d63226:3:core:0",
"prompt_tokens": 5178,
"target_tokens": 206,
"trained_tokens": 206,
"total_t... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/29 (Mon) 02:59", "messages": "[REAL_USER]: I'm looking for some gift wrapping ideas. I have a bunch of unique gifts, like a beautiful handmade candle and a set of artisanal soaps I got from a holiday market in my neighborhood on December 15th, and I want to make sure ... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Family: Has an aunt; preparing a personalized Christmas/holiday gift for her (handmade candle and artisanal soaps). Aunt's favorite scents: lavender and vanilla.\nGift/craft preferences: Values handmade, natura... | {
"row_id": "v5:4:run26-memory-review-checklist:1192316e__34__core:negative",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "hard-negative",
"error_type": 4,
"run": "run26-memory-review-checklist",
"trajectory": "runs/run26-memory-review-checklist/1192316e__34__cor... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
50yo Mumbai entrepreneur, CEO Grailzy (watch auction), co-founder Comparative (no-code analytics). NYC-based B2B sales/outreach for 6sense, writes for CRO/VP SaaS audiences. Tech: Spring Cloud, GraphQL, Angular, TensorFlow.js, Docker, Kubernetes.
Marathon runner (Chicago 4:10, trains 4-5x/week, 20-25mi/we... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The ukulele workshop was on 15 May 2023. The five-day New York City family trip ran from 24 May to 29 May 2023, so using the trip’s completion/return date, the interval from 15 May to 29 May is 14 days.","answer":"14 days","revisit_agent":"none","feedback":""}}
</t... | {
"row_id": "v5:12:run34ds-synthetic-date-arithmetic:e25c3b8d__0a32abd47c:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 12,
"run": "run34ds-synthetic-date-arithmetic",
"trajectory": "runs/run34ds-synt... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Rutgers student, Chinese immigrant parents. Lives within 4hrs of Outer Banks, NC. Morning person doing 9-5 construction law SEO content for Australian markets (NSW Building Act, payment disputes). Recurring meetings: Mon 10am, Thu 2pm. Calendar: 2-3 90-min focus blocks daily.
Entrepreneur: Nero men's skin... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000021","m000015","m000519","m000014"],"revisit_agent":"memory","feedback":"The question asks which specific financial aid options the user learned about for Rutgers study abroad on 2023-05-20. Repeated searches (financial aid, scholarships, grants, FAFSA, loans... | {
"row_id": "v5:10:run29ds-distractor-core:aae3761f__bea7dd6c01:1:4",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 97,
"source": "callback-decision",
"error_type": 10,
"run": "run29ds-distractor-core",
"trajectory": "runs/run29ds-distractor-core/aae... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Property appraisal/real estate professional. PropSci University, Georgetown political science background. San Diego 92130, moving Edinburgh for Master's (visa approved 5/22/23). Currently Nagpur, India.
Japanese-American filmmaker creating short film for LAAPFF. Inspired by Miyazaki, García Márquez. Tradi... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000597","m000485"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:15:run29ds-unanswerable:4bc144e2__73c71d2c24:3:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 15,
"run": "run29ds-unanswerable",
"trajectory": "runs/run29ds-unanswerable/4bc144e2__73c7... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Property appraisal/real estate professional. PropSci University, Georgetown political science background. San Diego 92130, moving Edinburgh for Master's (visa approved 5/22/23). Currently Nagpur, India.
Japanese-American filmmaker creating short film for LAAPFF. Inspired by Miyazaki, García Márquez. Tradi... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000415","m000438","m000275","m000109","m000413"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:4bc144e2:synthetic:membuilder_flat_after_v1:sft:4bc144e2:S29:72df23d381d94d99a65da6fef7c86bbb:5:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:4bc144e2:... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Marketing Specialist at XYZ Corporation (2+ years), reports to Rachel. Python, Pandas, Matplotlib, Seaborn. Planning data analytics certification.
Construction lawyer (NSW): building regulations, contracts, homeowner protection, educational content.
Full-stack developer: React, JavaScript, Node.js, Mongo... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The user-requested EPIC entries are m000418 for exlibrisgroup.com, m000422 for Marketo.com, and m000420 for Wiley.com, all adding Hindi, Hebrew, and Arabic with TMS and CMS integration. The Airbnb entry says the assistant generated an EPIC for Airbnb.com, not that ... | {
"row_id": "v5:14:run29ds-listing:6ade9755__36b74c8da4:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 401,
"source": "callback-decision",
"error_type": 14,
"run": "run29ds-listing",
"trajectory": "runs/run29ds-listing/6ade9755__36b74c8da4.json",
... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
LA-based social media marketing professional (~500 Twitter followers) seeking postdoctoral position. Background: Python/ML, SEO/digital marketing, content strategy. Frequent business traveler (5 trips/month). Developing RealEstateMall platform (Three.js, React).
Personality: Values persistence, self-compa... | <tool_call>
{"name":"search_memory","arguments":{"query":"shopping for sandals summer 2023 May","top_k":5,"banks":["episodic"]}}
</tool_call> | {
"row_id": "v5:v3:train:7024f17c:synthetic:membuilder_flat_after_v1:sft:7024f17c:S16:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:3",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7024f17c:... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Two independent entries (m000029 episodic and m000034 semantic, both from 21 May 2023) state the user planned to set up lists for To-Do, In Progress, and Done on their new Trello board. The procedural entry m000040 lists four stages including 'Ideas', but that is a... | {
"row_id": "v5:8:run18-hybrid-evidence-drop:91b15a6e__096f3be4ff:4:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-drop/9... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Female healthcare professional preparing for clinical informatics board exam. Works digital marketing 40-50 hrs/week, BA Business/Marketing UC Berkeley. Seeking Content Marketing Strategist role. Building AI-powered digital business. Military background using after-action review methodology.
Family: Siste... | <tool_call>
{"name":"search_memory","arguments":{"query":"jersey shadow box case size","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:c18a7dc8:synthetic:membuilder_flat_after_v1:sft:c18a7dc8:S38:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:c18a7dc8:... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/19 (Fri) 01:21", "messages": "[REAL_USER]: What files and configuration would I need to set up a docker compose server with\nNginx, Letsencrypt certbot, vhost, PHP 8.1 and MariaDB?\n[AI_ASSISTANT]: To set up a Docker Compose server with Nginx, Letsencrypt certbot, vho... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Technical Skills: Works with Docker, Docker Compose, server infrastructure. Experienced with Nginx web server configuration, PHP 8.1, MariaDB databases, Let's Encrypt SSL certificates (certbot). Understands vho... | {
"row_id": "v5:v3:train:gpt4_e061b84f:12:core:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:gpt4_e061b84f:12:core:0",
"prompt_tokens": 3734,
"target_tokens": 137,
"trained_tokens": 137... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000244","m000245","m000423","m000015","m000247"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_b5700ca0:synthetic:membuilder_flat_after_v1:sft:gpt4_b5700ca0:S24:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"search_memory","arguments":{"query":"volunteer opportunities high school students conservation community development organization","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_b5700ca0:synthetic:membuilder_flat_after_v1:sft:gpt4_b5700ca0:S38:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/08/28 (Mon) 01:48", "messages": "[REAL_USER]: I'm looking for some new board game recommendations. I recently played Ticket to Ride with my sister two weeks ago and we both loved it, so I'm looking for something similar. Do you have any suggestions?\n[AI_ASSISTANT]: Tick... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Board games: played Ticket to Ride with his sister around 2023-08-14 and both loved it; played Carcassonne with his dad on Sunday 2023-08-27, winning both games back-to-back. Prefers \"chill\" games that are ea... | {
"row_id": "v5:12:run37-shard0of4-synthetic-date-arithmetic:4b24c848__36b3cbc501:2:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 12,
"run": "run37-shard0of4-synthetic-date-arithmetic",
"trajectory": "run... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Software engineer, 5+ years experience. Former Research Officer at ADD Korea (UAV software), Full-Stack Developer at Conalog (cloud-native SaaS, IoT). Team leader applying for Palantir Forward Deployed Software Engineer role. Values diverse experiences, complex problem-solving, societal impact, teamwork, r... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000338","m000335","m000336","m000340","m000334"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_8279ba03:synthetic:membuilder_flat_after_v1:sft:gpt4_8279ba03:S30:72df23d381d94d99a65da6fef7c86bbb:5:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train... |
Manage MemBuilder procedural memory.
You are the Procedural Memory Manager. Manage step-by-step processes, workflows, and instructions.
Procedural Memory contains how-to guides, step-by-step instructions, or processes the user might follow.
Each procedural memory entry MUST include:
(a) entry_type: Type of procedure ... | {"bank": "procedural", "conversation_timestamp": "2023/05/29 (Mon) 08:08", "messages": "[REAL_USER]: I'm looking for some new breakfast ideas. I've been stuck in a rut lately and want to try something different. By the way, I had a great breakfast on April 10th - I made a dish with sweet potatoes, bell peppers, and scr... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"procedural","operations":[{"action":"ADD","memory":"Assembling roasted sweet potato, spinach, and egg breakfast burrito | Steps: 1. Add scrambled eggs, roasted sweet potatoes (diced or sliced bite-sized), and sautéed spinach and garlic to a tortilla. 2. Sprinkle ... | {
"row_id": "v5:2:run29ds-value-drop:4bc144e2__f7244d8979:4:3",
"row_role": "procedural",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 2,
"run": "run29ds-value-drop",
"trajectory": "runs/run29ds-value-drop/4bc144e2__f7244d8979.... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User Profile:
Name: Aman Ayaz
Contact: 647-717-5780, aman.ayaz44@gmail.com
Location: Tokyo, Japan (Roppongi, since Nov 2022)
Education & Career:
- 3rd-year Electrical Engineering, Toronto Metropolitan University
- MBA U Michigan (2021), Harvard CS50 (2021)
- Toronto Met Formula Racing (electrical drivetr... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000243","m000248"],"revisit_agent":"memory","feedback":"The memory records the electrician website project ('מארק - חשמלאי מוסמך' / Mark - Licensed Electrician) and notes the HTML contained a contact phone number, but no entry records the actual phone number dig... | {
"row_id": "v5:2:run25-value-drop:0862e8bf__afc871ecb2:1:4",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 77,
"source": "callback-decision",
"error_type": 2,
"run": "run25-value-drop",
"trajectory": "runs/run25-value-drop/0862e8bf__afc871ecb2.json"... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Female, 32 (Feb 15, 1991), California. Sister 35, niece 10. Strong Christian faith integrated with fitness (1 Cor 6:19).
Social media marketing/digital advertising pro (5 years): Facebook Ads, optimization, data analysis in R. Researches identity graphs, ML. Runs manifestation/spiritual alignment business... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000002","m000004","m000003","m000005","m000001"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:8:run18-hybrid-evidence-drop:d682f1a2__6cad5e54d7:3:4",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-dro... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
64-year-old male. BS Computer Science UCLA 2020 (GPA 3.6). From Arcadia, CA. Moved to Tokyo, Japan (August 20, 2023). Sister at Stanford studying biology. Mixed Indian-Pakistani heritage (mother from Rajasthan, father from Karachi).
Professional: Construction lawyer in NSW, Australia (uses Australian Engl... | <tool_call>
{"name":"search_memory","arguments":{"query":"Medellín market Colombia trip visit","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_372c3eed_abs:synthetic:membuilder_flat_after_v1:sft:gpt4_372c3eed_abs:S2:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id":... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
64-year-old male. BS Computer Science UCLA 2020 (GPA 3.6). From Arcadia, CA. Moved to Tokyo, Japan (August 20, 2023). Sister at Stanford studying biology. Mixed Indian-Pakistani heritage (mother from Rajasthan, father from Karachi).
Professional: Construction lawyer in NSW, Australia (uses Australian Engl... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Entries m000251 through m000256 all describe the user brainstorming Bonus Cards (spell, familiar, trinket themes) for their card game, and each is sourced from the session on 2023/08/20 (Sun) 17:06. No other date appears for Bonus Card design work.","answer":"20 Au... | {
"row_id": "v5:12:run30ds-temporal-reasoning:gpt4_372c3eed_abs__4ae5230e73:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 12,
"run": "run30ds-temporal-reasoning",
"trajectory": "runs/run30ds-temporal-re... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Female healthcare professional preparing for clinical informatics board exam. Works digital marketing 40-50 hrs/week, BA Business/Marketing UC Berkeley. Seeking Content Marketing Strategist role. Building AI-powered digital business. Military background using after-action review methodology.
Family: Siste... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000219"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:7:run29ds-retrieval-gap:c18a7dc8__482d2feddf:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 7,
"run": "run29ds-retrieval-gap",
"trajectory": "runs/run29ds-retrieval-gap/c18a7dc8__482d2... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/06/01 (Thu) 17:50", "messages": "[REAL_USER]: I'm considering entering a local art competition and I want to know more about the sculpture category. Can you give me some general tips on what judges usually look for in a sculpture piece? By the way, I've been taking sculp... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Art (sculpture, continued): Planning a series of small, intricate sculptures combining found objects with epoxy resin, experimenting with textures and effects; also interested in mixed media art. Chosen approac... | {
"row_id": "v5:15:run29ds-unanswerable:3ba21379__ba8601a7fd:2:1",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 15,
"run": "run29ds-unanswerable",
"trajectory": "runs/run29ds-unanswerable/3ba21379__ba8601a7f... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Emily Rose Thompson (post-divorce). Founder DesignSpark (Jan 15), creates Brand Style Guide Templates. Builds remote teams prioritizing emotional intelligence. Thesis: social media's mental health impact.
Projects: "Black Community Impact" gamification, Win'it e-commerce. Designs APIs/databases. Developin... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Multiple retrieved entries state the Representative level requires a $99 monthly subscription fee.","answer":"$99 per month","revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:1e043500:synthetic:membuilder_flat_after_v1:sft:1e043500:S18:72df23d381d94d99a65da6fef7c86bbb:2:answer:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:1e043500:synthe... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Sarah (#WriteWithSarah), ML researcher at leading tech firm, co-authored top-tier paper, TAing undergrad ML. Research: Visual Assistive Technology (YOLO/MobileNets), cultural heritage preservation, healthcare tech. SEO/content strategist for legal content. Prompt engineering consultant specializing in AI p... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for the three main categories of visual assistive technology devices in the user's presentation chapter. Entry m000263, from the same session on 24 May 2023, directly states the chapter on Visual Assistive Technology includes three main categories... | {
"row_id": "v5:6:run32-stale-core:88432d0a__4e2e35d779:4:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 6,
"run": "run32-stale-core",
"trajectory": "runs/run32-stale-core/88432d0a__4e2e35d779.json",
"... |
Manage MemBuilder procedural memory.
You are the Procedural Memory Manager. Manage step-by-step processes, workflows, and instructions.
Procedural Memory contains how-to guides, step-by-step instructions, or processes the user might follow.
Each procedural memory entry MUST include:
(a) entry_type: Type of procedure ... | {"bank": "procedural", "conversation_timestamp": "2023/05/20 (Sat) 15:48", "messages": "[REAL_USER]: What are some common myths about foot health that people should be aware of?\n[AI_ASSISTANT]: 1. MYTH: Foot pain is normal.\nFACT: Foot pain is not normal and should not be considered a part of aging. It is an indicatio... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"procedural","operations":[{"action":"ADD","memory":"How to maintain healthy feet | Steps: 1. Wear comfortable shoes that fit well, and avoid wearing high heels for extended periods of time. 2. Do regular foot exercises and stretches to maintain foot strength and ... | {
"row_id": "v5:v3:train:73d42213:2:procedural:0",
"row_role": "procedural",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:73d42213:2:procedural:0",
"prompt_tokens": 3291,
"target_tokens": 646,
"trained_tokens... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
LA resident. Fantasy writer (MK/Wukong, dialogue-driven), studies ecocriticism. Web/software developer (React, Python, Django, Rails) on real estate portal with ML models. Provides copywriting/marketing.
Marathon runner (April 15 PB), training for sprint triathlon. Yoga 2x/week, weekend cyclist. Landscape... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Entries m000526 and m000528 both state that Carmella's 2018 SmackDown Women's Championship win via Money in the Bank cash-in was historic because she was the first woman to successfully cash in the Money in the Bank contract, marking a significant moment for women'... | {
"row_id": "v5:8:run18-hybrid-evidence-drop:7e974930__12a3f209b6:4:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-drop/7... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
**Identity:** US-based secular compatibilist (last name: Thompson); values self-awareness, critical thinking. LGBTQ+ ally, gender equality advocate.
**Professional:** University international exchange department (student support programs). Background: real estate, social media. Skills: website/SEO, traini... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000444","m000434","m000462","m000436","m000437"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:10:run29ds-distractor-core:00ca467f__1543248b31:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 10,
"run": "run29ds-distractor-core",
"trajectory": "runs/run29ds-distractor-core/00ca467... |
Manage MemBuilder core memory.
You are the Core Memory Manager. Your role is to analyze user messages and extract fundamental information about the user that will be beneficial in future conversations.
Current Core Memory (Human Block):
[see input field: current_core_memory]
Character Usage: [see input field: core_usa... | {"bank": "core", "conversation_timestamp": "2023/05/23 (Tue) 16:07", "messages": "[REAL_USER]: can you assess the strong points and areas for improvement ifor the research proposal of an undergraduate student below, focusing on paragraph cohesion, sentence balance, flow and spelling and grammar accuracy. if possible ad... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"core","operations":[{"operation":"APPEND","content":"Academic Assessment: On 2023-05-23 the user assessed an undergraduate student's (Dune Blot, Biomedical Sciences, AY 2022-23) research proposal on endometriosis risk and gene variants (GWAS + exome sequencing); ... | {
"row_id": "v5:12:run30ds-temporal-reasoning:gpt4_b4a80587__49c4381223:2:0",
"row_role": "core",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 12,
"run": "run30ds-temporal-reasoning",
"trajectory": "runs/run30ds-temporal-reason... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
**Identity:** US-based secular compatibilist (last name: Thompson); values self-awareness, critical thinking. LGBTQ+ ally, gender equality advocate.
**Professional:** University international exchange department (student support programs). Background: real estate, social media. Skills: website/SEO, traini... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for two craft supplies used to add decorative touches. Entries m000196 and m000197, both from the session at 2023/03/27 10:45 (matching the question date), describe the user using Copic markers for colorful touches and embroidery floss/threads for... | {
"row_id": "v5:15:run29ds-unanswerable:00ca467f__6324f3ceff:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 574,
"source": "callback-decision",
"error_type": 15,
"run": "run29ds-unanswerable",
"trajectory": "runs/run29ds-unanswerable/00ca467f__6324... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000111","m000105","m000109","m000110","m000103","m000148"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:7:run18-hybrid-retrieval-gap:91b15a6e__1baba26dd2:2:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 7,
"run": "run18-hybrid-retrieval-gap",
"trajectory": "runs/run18-hybrid-retrieval-ga... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Business Owner: Operates Indoshep, warehouse-only Indonesian grocery store in Tasmania, Australia (opened April 2023). Developing TikTok marketing and SEO strategy.
Education & Career: Studying IT for Managers. Considering Ph.D. in education technology or breath-based disease detection. Quit marketing fir... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000210","m000214","m000212","m000209","m000206"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:12:run30ds-temporal-reasoning:1a8a66a6__c90dee47fc:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 12,
"run": "run30ds-temporal-reasoning",
"trajectory": "runs/run30ds-temporal-reasonin... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
LA-based social media marketing professional (~500 Twitter followers) seeking postdoctoral position. Background: Python/ML, SEO/digital marketing, content strategy. Frequent business traveler (5 trips/month). Developing RealEstateMall platform (Three.js, React).
Personality: Values persistence, self-compa... | <tool_call>
{"name":"search_memory","arguments":{"query":"lunch meal prep healthy eating habits diet","top_k":5,"banks":["episodic","semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:v3:train:7024f17c:synthetic:membuilder_flat_after_v1:sft:7024f17c:S26:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7024f17c:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
QA Automation Test Engineer at Wahoo Fitness (remote, Minneapolis, Oct 2021, promoted Nov 2022). Specializes in mobile apps, automation, embedded devices. Skills: Python, pytest, BDD, CI/CD, Jenkins, REST API. Leads bike computer automation. Interested in Sr. QA role at 1Password.
Genealogy researcher: 19... | <tool_call>
{"name":"search_memory","arguments":{"query":"pregnancy mood swings management deep breathing meditation prenatal yoga","top_k":8,"banks":["episodic","semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:8:run18-hybrid-evidence-drop:945e3d21__74b077768f:3:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidence-dro... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Rutgers student, Chinese immigrant parents. Lives within 4hrs of Outer Banks, NC. Morning person doing 9-5 construction law SEO content for Australian markets (NSW Building Act, payment disputes). Recurring meetings: Mon 10am, Thu 2pm. Calendar: 2-3 90-min focus blocks daily.
Entrepreneur: Nero men's skin... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Core memory states the area rug arrived May 20, 2023 and the piano overview expansion was requested May 24, 2023. From 20 May to 24 May is 4 days.","answer":"4 days","revisit_agent":"retrieval","feedback":"No retrieved entry confirms Core's dates for the area rug a... | {
"row_id": "v5:6:run32-stale-core:aae3761f__0eb5f3de85:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 221,
"source": "callback-decision",
"error_type": 6,
"run": "run32-stale-core",
"trajectory": "runs/run32-stale-core/aae3761f__0eb5f3de85.json",
... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Telugu culture organizer (NATS, America Telugu Sambaralu). Graduate student: L2 Motivational Self-System/second language acquisition (TESOL). NYC resident. Marketing professional, considered for Senior Marketing Specialist (manager: Rachel).
45-year DJ: Technics 1200, worldwide clubs, underground music. V... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for advice on what to focus on next in game development. The retrieved entries show the user already gathered learning resources for Unity (Unity Learn, Udemy, YouTube channels like Brackeys, Code Monkey, Game Dev Guide, Sykoo, Jayanam) and for pi... | {
"row_id": "v5:16:run30ds-preference:ce6d2d27__289af06c29:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 1085,
"source": "callback-decision",
"error_type": 16,
"run": "run30ds-preference",
"trajectory": "runs/run30ds-preference/ce6d2d27__289af06c2... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
32yo Italian-speaking construction lawyer (homeowner rights, contracts). Studies philosophy, theology, Vatican II, existentialism, science-religion dialogue. Familiar with Freud, Frankl, Jung, Nietzsche, Marx, Dawkins, Gandhi, MLK Jr. Birthday: Feb 15.
Finances: $4,500/month income, $3,500 expenses. Debts... | <tool_call>
{"name":"search_memory","arguments":{"query":"anthropology definition learned","top_k":5,"banks":["semantic","episodic"]}}
</tool_call> | {
"row_id": "v5:v3:train:89941a94:synthetic:membuilder_flat_after_v1:sft:89941a94:S1:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:89941a94:s... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
50yo Mumbai entrepreneur, CEO Grailzy (watch auction), co-founder Comparative (no-code analytics). NYC-based B2B sales/outreach for 6sense, writes for CRO/VP SaaS audiences. Tech: Spring Cloud, GraphQL, Angular, TensorFlow.js, Docker, Kubernetes.
Marathon runner (Chicago 4:10, trains 4-5x/week, 20-25mi/we... | <tool_call>
{"name":"search_memory","arguments":{"query":"New Zealand Australia trip duration weeks","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:e25c3b8d:synthetic:membuilder_flat_after_v1:sft:e25c3b8d:S20:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:e25c3b8d:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000404","m000406","m000405","m000407"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:8:run18-hybrid-evidence-drop:gpt4_b5700ca0__49ce8ad496:3:4",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 8,
"run": "run18-hybrid-evidence-drop",
"trajectory": "runs/run18-hybrid-evidenc... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
64-year-old male. BS Computer Science UCLA 2020 (GPA 3.6). From Arcadia, CA. Moved to Tokyo, Japan (August 20, 2023). Sister at Stanford studying biology. Mixed Indian-Pakistani heritage (mother from Rajasthan, father from Karachi).
Professional: Construction lawyer in NSW, Australia (uses Australian Engl... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000001","m000007","m000139","m000149","m000143","m000361","m000356","m000150"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:5:run29ds-subject-drop:gpt4_372c3eed_abs__1a094f1aa1:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "hard-negative",
"error_type": 5,
"run": "run29ds-subject-drop",
"trajectory": "runs/run29ds-subject-drop/gpt4_372c... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User Profile:
Name: Nahom Mekonen (ID: 14917430)
Location: Harajuku, Tokyo (since ~March 2023)
Birthday: June
Career:
- Secondary school science teacher (secondary 1): hands-on inquiry-based projects, structured scientific method
- Fiverr data analyst: COVID-19/healthcare, time-series, Python
- Runs "Veg... | <tool_call>
{"name":"search_memory","arguments":{"query":"prophase I homologous chromosomes exchange genetic information process","top_k":5,"banks":["semantic","procedural"]}}
</tool_call> | {
"row_id": "v5:v3:train:2133c1b5:synthetic:membuilder_flat_after_v1:sft:2133c1b5:S1:72df23d381d94d99a65da6fef7c86bbb:3:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:2133c1b5:s... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
User Profile:
- English instructor at Texas community college (Distance Ed/Educational Technology); detail-oriented, methodical, data-driven
- Bilingual English/Spanish; learning Mandarin
- Lives on Indiana farm; moved to UK for Edinburgh study abroad (visa approved May 27, 2023)
Family: Brother graduated... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":[],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:5:run31-subject-drop:d01c6aa8__97333bb8ab:3:0",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 5,
"run": "run31-subject-drop",
"trajectory": "runs/run31-subject-drop/d01c6aa8__97333bb8ab.j... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Software engineer, 5+ years experience. Former Research Officer at ADD Korea (UAV software), Full-Stack Developer at Conalog (cloud-native SaaS, IoT). Team leader applying for Palantir Forward Deployed Software Engineer role. Values diverse experiences, complex problem-solving, societal impact, teamwork, r... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The question asks for the protagonist of 'The City of Dreaming Books.' Core memory says Hildegunst von Mythenmetz, but the retrieved entries m000433 and m000426 explicitly identify the protagonist as Optimus Yarnspinner, a young author who travels to Bookholm. The ... | {
"row_id": "v5:6:run29ds-stale-core:gpt4_8279ba03__c0a6bec0a9:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 6,
"run": "run29ds-stale-core",
"trajectory": "runs/run29ds-stale-core/gpt4_8279ba03__c0a6b... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Finance/asset management professional: treasury operations, collateral management, derivatives, regulatory compliance. Retail shop owner. Founder/leader of Codestates (talent transformation company).
Account Executive at OwnBackup (SaaS) since Feb 2021. 10+ years sales. President's Club 2022, Top Performe... | <tool_call>
{"name":"search_memory","arguments":{"query":"Gucci handbag purchase date","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:86b68151:synthetic:membuilder_flat_after_v1:sft:86b68151:S33:72df23d381d94d99a65da6fef7c86bbb:5:retrieval:1",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:86b68151:... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
**Identity:** US-based secular compatibilist (last name: Thompson); values self-awareness, critical thinking. LGBTQ+ ally, gender equality advocate.
**Professional:** University international exchange department (student support programs). Background: real estate, social media. Skills: website/SEO, traini... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000069","m000070","m000236","m000073","m000071"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:00ca467f:synthetic:membuilder_flat_after_v1:sft:00ca467f:S6:72df23d381d94d99a65da6fef7c86bbb:2:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:00ca467f:s... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Sarah (#WriteWithSarah), ML researcher at leading tech firm, co-authored top-tier paper, TAing undergrad ML. Research: Visual Assistive Technology (YOLO/MobileNets), cultural heritage preservation, healthcare tech. SEO/content strategist for legal content. Prompt engineering consultant specializing in AI p... | <tool_call>
{"name":"answer","arguments":{"reasoning":"Entry m000120 dates the disappointing sourdough bread recipe to 16 May 2023 (the Tuesday before the 21 May 2023 conversation; 16 May 2023 was indeed a Tuesday). For the whole wheat baguette, entry m000448 says \"last Saturday\" from the Tuesday 30 May 2023 conversa... | {
"row_id": "v5:12:run37-shard0of4-synthetic-date-arithmetic:88432d0a__49a35fd433:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 12,
"run": "run37-shard0of4-synthetic-date-arithmetic",
"trajectory": "r... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The retrieved entries from 8 March 2023 say the user is working on the backhand slice and has been experiencing trouble with consistency, with footwork and court positioning identified as areas to improve.","answer":"Consistency, especially footwork and court posit... | {
"row_id": "v5:5:run29ds-subject-drop:gpt4_b5700ca0__f35983cf54:3:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 5,
"run": "run29ds-subject-drop",
"trajectory": "runs/run29ds-subject-drop/gpt4_b5700ca0_... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
User: Business Admin grad (U Michigan 2015, Finance). Former farmer, email marketing/copywriting pro transitioning to data science/ML in finance via Coursera. Learning Python, ML libraries (scikit-learn, TensorFlow, pandas, numpy, nltk). Targeting Bloomberg, Goldman Sachs, JPMorgan, Citigroup. Technical: D... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The relevant episodic memory states the user asked how pulsar timing measurements were used to confirm Einstein's theory of general relativity regarding the mass of Sirius.","answer":"How pulsar timing measurements were used to confirm Einstein's theory of general ... | {
"row_id": "v5:v3:train:91b15a6e:synthetic:membuilder_flat_after_v1:sft:91b15a6e:S0:72df23d381d94d99a65da6fef7c86bbb:1:answer:2",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:91b15a6e:synthet... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
LA resident. Fantasy writer (MK/Wukong, dialogue-driven), studies ecocriticism. Web/software developer (React, Python, Django, Rails) on real estate portal with ML models. Provides copywriting/marketing.
Marathon runner (April 15 PB), training for sprint triathlon. Yoga 2x/week, weekend cyclist. Landscape... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000300","m000297","m000296","m000301","m000299"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:7e974930:synthetic:membuilder_flat_after_v1:sft:7e974930:S28:72df23d381d94d99a65da6fef7c86bbb:1:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7e974930:... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Senior Marketing Specialist at LumenPay (DeFi/blockchain), promoted June 1st. Preparing IoT Expo Chicago with Alex, Rachel, Michael. Pursuing MBA (Harvard, Stanford, Wharton, INSEAD, MIT Sloan), studying for GMAT. Master's thesis: NLP model analyzing student essays (10,000+ dataset). Works in behavioral re... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The entries from 6 April 2023 say the user was considering a portable payment processor or mobile card reader to make payment easier and track sales, and planned to have small bills and coins for cash transactions. The QR code/mobile wallet entry is dated 12 May 20... | {
"row_id": "v5:9:run29ds-distractor-rank:gpt4_b5700ca0__29b93eca05:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 529,
"source": "callback-decision",
"error_type": 9,
"run": "run29ds-distractor-rank",
"trajectory": "runs/run29ds-distractor-rank/gp... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
San Francisco resident (May 2023), later South Australia, now Chicago (May 2023+). Pets: Max (dog), Luna (cat). Vegetarian/vegan, mild foods, managing diabetes. Close with 86-year-old grandmother. Formally diagnosed Autistic, registered with NDIS. Severe anxiety, especially at night and uncertain situation... | <tool_call>
{"name":"finish","arguments":{"ranked_ids":["m000385","m000156","m000387","m000386","m000166"],"revisit_agent":"none","feedback":""}}
</tool_call> | {
"row_id": "v5:v3:train:7527f7e2:synthetic:membuilder_flat_after_v1:sft:7527f7e2:S32:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:7527f7e2:... |
Manage MemBuilder semantic memory.
You are the Semantic Memory Manager. Manage conceptual knowledge about people, places, objects, and concepts.
Semantic Memory holds general knowledge, concepts, definitions, and facts. It is the storehouse of abstract understanding about the world.
IMPORTANT: ONLY save NEW concepts ... | {"bank": "semantic", "conversation_timestamp": "2023/05/29 (Mon) 21:17", "messages": "[REAL_USER]: generate a list for: Smoothie Bowl, Avocado Toast,\n\ninstructions without proportions,\nyou HAVE to include the calories\n\nuse this structure: \n\nDish Name: Dish\nIngredients: \n1. ingredient (ammount) (unit)\n2. ingre... | <tool_call>
{"name":"manage_memory","arguments":{"bank":"semantic","operations":[{"action":"ADD","memory":"Avocado Toast recipe (user-requested): 5-minute breakfast with ~300 calories | Details: User requested a recipe list including Avocado Toast with calories. Ingredients: 1 avocado, 2 slices whole wheat bread, salt ... | {
"row_id": "v5:1:run18-hybrid-memory-gap:f523d9fe__9ff89603cd:2:2",
"row_role": "semantic",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": null,
"source": "callback-redo",
"error_type": 1,
"run": "run18-hybrid-memory-gap",
"trajectory": "runs/run18-hybrid-memory-gap/f523d9fe... |
You search a memory built from a user's earlier conversations with an assistant, to collect
the evidence needed to answer a question about the user. Each memory entry has a memory_id, the dates of the
sessions it came from, and its content. The question date is "now" for the question.
Search rules:
- Core below is con... | Core memory:
Senior Software Engineer at TechCorp (agricultural tech), aspiring Technical Lead/Architect. Expertise: data science, ML, Python, AI, cloud computing, Docker, WebSockets, WebRTC, Node.js, React. Building video chat app. Owns CreativeSpark Marketing agency, YouTube/Instagram creator launching Patreon "Marke... | <tool_call>
{"name":"search_memory","arguments":{"query":"yoga soccer weekly activities training","top_k":5,"banks":["episodic","semantic"]}}
</tool_call> | {
"row_id": "v5:v3:train:gpt4_2ba83207:synthetic:membuilder_flat_after_v1:sft:gpt4_2ba83207:S6:72df23d381d94d99a65da6fef7c86bbb:4:retrieval:2",
"row_role": "retrieval",
"format_protocol": "memory_banks_tool_call_v5_callback",
"source": "v3",
"error_type": null,
"loss_from_char": null,
"v3_row_id": "train:... |
You answer a question about a user from evidence collected from a memory of the user's earlier
conversations with an assistant: the Core memory and the retrieved memory entries listed below. Each entry has a
memory_id, the dates of the sessions it came from, and its content. The question date is "now" for the question.... | Core memory:
Business Owner: Operates Indoshep, warehouse-only Indonesian grocery store in Tasmania, Australia (opened April 2023). Developing TikTok marketing and SEO strategy.
Education & Career: Studying IT for Managers. Considering Ph.D. in education technology or breath-based disease detection. Quit marketing fir... | <tool_call>
{"name":"answer","arguments":{"reasoning":"The user bought the National Geographic issue on 20 March 2023 (m000549: 'User bought the National Geographic issue' at a local bookstore). The user learned about Kurt Busch's philanthropic efforts on 1 April 2023 15:25 (m000362). From 20 March 2023 to 1 April 2023... | {
"row_id": "v5:9:run32-distractor-rank:1a8a66a6__710e6d3439:2:0",
"row_role": "answer",
"format_protocol": "memory_banks_tool_call_v5_callback",
"loss_from_char": 392,
"source": "callback-decision",
"error_type": 9,
"run": "run32-distractor-rank",
"trajectory": "runs/run32-distractor-rank/1a8a66a6__710... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.