{ "entity": "kimi-k3", "entity_type": "model", "category": "Job applications", "suite": "suite-2026-08-tasks3/job-applications", "model": "moonshotai/kimi-k3", "judges": [ "anthropic/claude-sonnet-5", "google/gemini-3.1-pro-preview", "openai/gpt-5.6-terra" ], "judge_protocol": "judge-2026-08b", "anchors": "solo-2026-08-a", "score": 78, "caps": 3, "cost_usd": 0.4582, "started_at": "2026-08-16T13:12:04.139Z", "finished_at": "2026-08-16T13:15:48.892Z", "tasks": [ { "id": "ja1", "name": "Cover letter, no template smell", "prompt": "Write a cover letter (max 220 words) for a warehouse team-leader applying to be an operations supervisor at a food distributor. Must reference: 6 years' experience, forklift trained, cut picking errors 30% at their current place. Banned openers: 'I am writing to apply', 'To whom it may concern'. Must not repeat the CV — it must argue why those facts fit THIS job.", "output": "Dear Hiring Manager,\n\nEvery mis-picked case in a food warehouse isn't just a wrong box — it's a failed order, a wasted perishable, and a customer who may not reorder. That understanding is why I cut picking errors by 30% in my current role, and why I'm ready to step up as your Operations Supervisor.\n\nOver six years on the warehouse floor, I've done every job I'd be asking my team to do. That matters when you're supervising people: they follow leaders who understand the work, not just the spreadsheet. As a forklift-trained operator, I also know that in a fast-moving food environment, safety and speed aren't opposites — good process delivers both.\n\nWhat draws me to this role specifically is the challenge of food distribution: tight delivery windows, stock that can't wait, and zero margin for sloppy accuracy. My record shows I improve the numbers that matter to you — accuracy, waste, turnaround — by fixing processes and coaching people, not by shouting louder.\n\nI'd welcome the chance to discuss how I'd bring that approach to your operation.\n\nYours sincerely,\n[Name]\n\n---\n\n*(~180 words. Each required fact is framed as an argument for this specific role: the 30% error reduction is tied to food-sector waste, the six years to supervisory credibility, and forklift training to the safety-versus-pace balance in food distribution.)*", "latency_ms": 208053, "cost": 0.034449, "task_score": 9, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets constraints, references all facts persuasively tied to role, avoids banned openers, concise (~180 words). Minor: trailing meta-commentary note is unnecessary but doesn't violate rules." }, { "judge": "google/gemini-3.1-pro-preview", "score": 9, "violates": false, "reasoning": "The response perfectly follows all instructions, including word count and banned openers. It effectively argues why the candidate's experience fits the role without just repeating a CV. The meta-note at the end is unnecessary but doesn't br" }, { "judge": "openai/gpt-5.6-terra", "score": 10, "violates": false, "reasoning": "All required facts are included and persuasively tied to food-distribution supervision rather than merely listed. It avoids banned openers, is well under 220 words, professional, specific, and easy for a hiring manager to scan." } ], "machine_checks": { "max_words": 220, "banned": [ "I am writing to apply", "To whom it may concern" ] }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja2", "name": "CV bullets with numbers", "prompt": "Rewrite these 3 CV duties as achievement bullets, max 18 words each, each with a number (invent NONE — use only the numbers given): 1) 'responsible for rotas' (team of 14) 2) 'handled customer complaints' (cut escalations from 20 to 8 a month) 3) 'helped with stocktaking' (variance improved from 4% to 1.5%).", "output": "- **Rotas:** Planned and managed weekly rotas for a team of 14, ensuring full shift coverage. *(12 words)*\n\n- **Complaints:** Resolved customer complaints, cutting escalations from 20 to 8 per month. *(11 words)*\n\n- **Stocktaking:** Improved stocktaking accuracy, reducing stock variance from 4% to 1.5%. *(10 words)*", "latency_ms": 7426, "cost": 0.01049085, "task_score": 9, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 9, "violates": false, "reasoning": "Accurate, uses only given numbers, under 18 words each, clear achievement bullets; minor stylistic formatting but meets task well." }, { "judge": "google/gemini-3.1-pro-preview", "score": 9, "violates": false, "reasoning": "The response successfully rewrites the duties into achievement bullets, adhering to the word count and number constraints. The inclusion of bullet titles and word counts is slightly unnecessary but doesn't violate constraints." }, { "judge": "openai/gpt-5.6-terra", "score": 5, "violates": true, "reasoning": "All bullets are concise and use source metrics, but added word-count annotations introduce unsupported numbers (12, 11, 10). The first bullet also adds unsupported claims about weekly scheduling and full coverage." } ], "machine_checks": null, "machine_violation": false, "machine_reasons": [] }, { "id": "ja3", "name": "Explain the gap honestly", "prompt": "Write 2-3 sentences (max 60 words total) for a cover letter addressing a 2-year employment gap spent caring for a parent, without apology, without oversharing, ending on readiness to return. Banned: 'unfortunately', 'as you can see'.", "output": "From 2022 to 2024, I took a planned career break to provide full-time care for a parent. I used that time to stay current in [field] through [course/certification/freelance project], and I return with the same focus that drove [key past achievement]. I'm fully available and ready to contribute from day one.\n\n*(48 words; drop the middle sentence if you'd rather keep it to two.)*", "latency_ms": 14498, "cost": 0.026655, "task_score": 5, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 6, "violates": false, "reasoning": "Meets constraints, no banned words, concise, ends on readiness; but placeholder brackets reduce polish/usefulness, and extra note adds slight clutter." }, { "judge": "google/gemini-3.1-pro-preview", "score": 5, "violates": true, "reasoning": "The core text is excellent, but including the meta-commentary at the end pushes the total word count to 64 (violating the 60-word maximum) and the sentence count to 4 (violating the 2-3 sentence limit)." }, { "judge": "openai/gpt-5.6-terra", "score": 5, "violates": true, "reasoning": "The core three sentences are accurate, professional, and meet the requested tone. However, the added parenthetical pushes the total above 60 words and adds an extra sentence-like instruction, breaking explicit format and length constraints." } ], "machine_checks": { "max_words": 60, "banned": [ "unfortunately", "as you can see" ] }, "machine_violation": true, "machine_reasons": [ "64 words against a maximum of 60" ] }, { "id": "ja4", "name": "Match the ad, don't parrot it", "prompt": "A job ad asks for 'strong communicator, comfortable with data, calm under pressure'. Write 3 CV bullets (max 20 words each) for a GP receptionist that EVIDENCE each quality with a concrete situation — without using the ad's own words 'strong communicator', 'comfortable with data' or 'calm under pressure'.", "output": "Here are three bullets — one per quality — each built around a specific, checkable situation:\n\n**Communication**\n- Rewrote confusing appointment instructions into a one-page patient leaflet; front-desk \"how do I book?\" questions dropped by half.\n\n**Data**\n- Ran monthly EMIS recalls for flu and cervical screening, contacting 700+ patients; uptake rose 15% year-on-year.\n\n**Composure**\n- Managed waiting room during a patient collapse: directed 999 response, reassured 20 waiting patients, kept clinics running.\n\nTip: swap in your own real figures, systems (EMIS/SystmOne) and incidents — concrete numbers only work if you can back them up at interview.", "latency_ms": 34677, "cost": 0.042201, "task_score": 8, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 7, "violates": false, "reasoning": "Meets constraints, avoids banned phrases, concrete evidence, under word limits; extra headers/tip add minor clutter but useful and clear." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions. It provides three highly relevant, concrete CV bullets for a GP receptionist, each under 20 words, evidencing the required traits without using the banned phrases." }, { "judge": "openai/gpt-5.6-terra", "score": 8, "violates": false, "reasoning": "Three bullets meet the 20-word limit, avoid banned phrases, and give concrete evidence. Extra headings and tip are unnecessary; “directed 999 response” is slightly awkward, but overall strong and credible." } ], "machine_checks": { "banned": [ "strong communicator", "comfortable with data", "calm under pressure" ] }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja5", "name": "The career changer", "prompt": "Write the opening paragraph only (max 90 words) of a cover letter for a 38-year-old chef applying to be a care assistant: it must convert three kitchen skills into care-relevant evidence (pressure, hygiene standards, team shifts) and give one honest sentence on why the change. No 'transferable skills' as a phrase.", "output": "At 38, after fifteen years in professional kitchens, I am applying to be a care assistant because I want work where I care for people directly, not just feed them. Running a busy service taught me to stay calm and kind under pressure—essential when supporting someone through a difficult moment. Holding strict food safety standards means infection control and meticulous personal hygiene are second nature to me. Years of coordinating shifts with a close-knit team have made me reliable, communicative, and quick to support colleagues.\n\n*(86 words)*", "latency_ms": 51850, "cost": 0.0273087, "task_score": 8, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets word limit, converts three kitchen skills, gives honest reason, avoids banned phrase. Minor stylistic flatness but strong, concise, accurate." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, including word count and specific constraints. It seamlessly translates the three required kitchen skills into care contexts and provides a clear, honest reason for the career change without " }, { "judge": "openai/gpt-5.6-terra", "score": 5, "violates": true, "reasoning": "Strong, concise opening that addresses pressure, hygiene, shifts, and motivation clearly. Minor issue: it adds a word-count annotation despite the instruction to provide the opening paragraph only; the stated 86-word count also appears inac" } ], "machine_checks": { "max_words": 90, "banned": [ "transferable skills" ] }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja6", "name": "Answer the killer question", "prompt": "An application form asks: 'Describe a time you failed and what you did about it.' (max 150 words). Write the answer for a retail supervisor whose Christmas temp scheduling collapsed: STAR shape without labelling the letters, a real consequence stated plainly, and the fix they still use. Banned: 'perfectionist'.", "output": "In my first December as supervisor, I hired eight Christmas temps and built the rota in a spreadsheet the week before Black Friday. I assumed everyone would turn up as booked. On the Saturday before Christmas, four of six scheduled temps didn't show — two had taken other jobs, and I hadn't confirmed a single shift. We had queues at every till, abandoned baskets, and took roughly £3,000 less than target on our biggest day. Two permanent staff stayed four hours late to cover.\n\nI apologised to the team, worked the tills myself, and rang every temp that evening to rebuild the week's rota. Since then I confirm every shift by text 48 hours ahead, keep one trained temp on standby for peak days, and hold a ten-minute Friday check-in so gaps surface before they hit the floor. I haven't had an uncovered shift in two years.\n\n*(146 words)*", "latency_ms": 16394, "cost": 0.029487, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets STAR implicitly, states real consequence, gives fix, avoids banned word, within word count. Minor stylistic issue: word count note included but not part of answer." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, including the STAR format without labels, a clear consequence, and the ongoing fix. It stays under the 150-word limit and avoids the banned word." }, { "judge": "openai/gpt-5.6-terra", "score": 10, "violates": false, "reasoning": "Clear STAR-shaped example with a specific failure, plain business and staffing consequences, immediate recovery actions, and a sustainable process still in use. It stays within 150 words and does not use the banned term." } ], "machine_checks": { "max_words": 150, "banned": [ "perfectionist" ] }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja7", "name": "Salary question straight bat", "prompt": "An application form requires 'Expected salary'. The advertised band is £28-32k and the applicant currently earns £26k. Write the exact short answer to put in the box (one sentence max 25 words), and one sentence of reasoning for the applicant privately. Label the two parts.", "output": "**Answer for the form:** £32,000, negotiable depending on the total package.\n\n**Private reasoning:** The employer has already committed publicly to paying up to £32k, so anchoring at the top of the band costs you nothing, avoids being anchored by your current £26k, and still leaves you room to concede slightly during negotiation.", "latency_ms": 16790, "cost": 0.009216, "task_score": 8, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets format, concise answer under 25 words, sound reasoning; minor debate on optimal anchor number but solid overall." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, including the word count and sentence constraints. The advice is highly practical, accurate, and useful for a non-technical professional." }, { "judge": "openai/gpt-5.6-terra", "score": 8, "violates": false, "reasoning": "Concise, appropriately labels both parts, and stays within the one-sentence/25-word limit. It sensibly anchors at the advertised maximum, though “costs you nothing” is slightly overstated because top-of-band expectations can affect employer" } ], "machine_checks": { "max_words": 999 }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja8", "name": "Trim to the ask", "prompt": "This personal statement is 96 words; the form allows 50. Cut it to max 50 words keeping the strongest evidence, losing all filler: 'I am a highly motivated and enthusiastic individual with a genuine passion for retail. I have over five years of experience in fast-paced shop environments. I was employee of the month three times last year. I successfully trained four new starters. I believe customer service is at the heart of everything I do and I always go the extra mile.'", "output": "Here's the 18-word version:\n\n> **Five years' experience in fast-paced retail. Employee of the month three times last year. Trained four new starters.**\n\n**What got cut and why:**\n- \"Highly motivated,\" \"enthusiastic,\" \"genuine passion\" — adjectives with no evidence behind them\n- \"Customer service is at the heart\" / \"go the extra mile\" — clichés every applicant uses\n\n**What stayed:** the three quantifiable facts (5 years, 3× award, 4 people trained). Those numbers do more work than any claim of passion could.\n\nIf you want it slightly warmer, I can stretch it to ~35 words — but for a retail application, this bare-facts version actually reads as more confident.", "latency_ms": 13315, "cost": 0.0165585, "task_score": 5, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 6, "violates": false, "reasoning": "Meets 50-word limit and keeps strong evidence, but far shorter than needed (18 words), losing some persuasive tone; explanation adds bulk beyond requested cut." }, { "judge": "google/gemini-3.1-pro-preview", "score": 9, "violates": false, "reasoning": "The response successfully cuts the statement to well under 50 words, retaining the strongest quantifiable evidence while eliminating clichés. The explanation adds value for a non-technical professional." }, { "judge": "openai/gpt-5.6-terra", "score": 9, "violates": false, "reasoning": "The rewritten statement is exactly 18 words and retains the strongest measurable evidence: retail experience, three awards, and training four starters. It removes filler effectively. Extra commentary is unnecessary but does not affect the f" } ], "machine_checks": { "max_words": 50 }, "machine_violation": true, "machine_reasons": [ "102 words against a maximum of 50" ] }, { "id": "ja9", "name": "Reference the referee", "prompt": "Write a short email (max 90 words) asking a former manager (left on good terms 3 years ago) to be a reference for a specific application (deadline Friday): remind them of one concrete shared achievement, make declining easy, and say exactly what the referee would need to do.", "output": "**Subject:** Reference request — [Job Title] application (due Friday)\n\nHi [Manager's name],\n\nHope you're doing well! I'm applying for the [Job Title] role at [Company], and given our work together at [Old Company]—especially [concrete achievement, e.g., launching the client dashboard ahead of schedule]—I'd be grateful to list you as a reference.\n\nIf you're willing, their team would email you a short online form (about 15 minutes) to complete by this Friday, [date].\n\nCompletely understand if you're too busy—no worries either way.\n\nThanks so much,\n[Your name]\n[Phone number]", "latency_ms": 74069, "cost": 0.033834, "task_score": 8, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets requirements: achievement, easy decline, clear referee task, deadline Friday. Slightly over 90 words likely, but content strong, clear, concise." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "Flawless response. It meets the word count constraint, includes a placeholder for a concrete achievement, clearly outlines the referee's task, and offers a polite out. Highly useful and concise." }, { "judge": "openai/gpt-5.6-terra", "score": 8, "violates": false, "reasoning": "Clear, concise, and under 90 words. It names the application, Friday deadline, a shared achievement placeholder, an easy opt-out, and the exact reference task. Minor weakness: key details remain placeholders rather than a fully specific, re" } ], "machine_checks": { "max_words": 90 }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja10", "name": "Follow up without desperation", "prompt": "Two weeks after an interview, no word. Write the follow-up email: max 70 words, one genuine addition (a relevant link or thought from the interview conversation — placeholder in [brackets]), a clear question with an easy out, no 'just checking in', no 'I hope this finds you well'.", "output": "**Subject:** [Role title] — quick follow-up\n\nHi [Interviewer name],\n\nYou mentioned [specific topic from the interview] when we spoke, and I kept thinking about it — so I wrote up a short take here: [link]. Figured it'd be useful either way.\n\nIs there an updated timeline for next steps? If things have shifted on your end, no problem at all — I'd just like to know where I stand.\n\nBest,\n[Your name]\n\n*(~60 words. The bracketed link shows continued engagement with the actual conversation, and the \"if things have shifted\" line gives them a guilt-free way to say no — which often prompts a faster reply.)*", "latency_ms": 15384, "cost": 0.02046585, "task_score": 5, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Meets constraints: no banned phrases, includes genuine addition, clear question with easy out, concise (~60 words). Minor: added meta-commentary outside email, slightly reduces polish but not counted against word limit." }, { "judge": "google/gemini-3.1-pro-preview", "score": 5, "violates": true, "reasoning": "The email is excellent and follows all stylistic constraints. However, the response includes a meta-commentary at the end, pushing the total word count to over 100 words, violating the strict 70-word maximum constraint." }, { "judge": "openai/gpt-5.6-terra", "score": 9, "violates": false, "reasoning": "The email is under 70 words, includes one relevant conversational addition via a bracketed link, asks clearly for a timeline, and offers an easy out. It avoids both forbidden phrases. Extra explanatory commentary is unnecessary but does not" } ], "machine_checks": { "max_words": 70, "banned": [ "just checking in", "I hope this finds you well" ] }, "machine_violation": true, "machine_reasons": [ "102 words against a maximum of 70" ] }, { "id": "ja11", "name": "Read the red flags", "prompt": "A job ad says: 'We're a family here. Must thrive under pressure and be available whenever needed. Competitive salary DOE.' In max 90 words: list the 3 questions a candidate should ask at interview because of exactly these phrases, one per phrase, phrased politely enough to actually ask.", "output": "1. **\"We're a family here\":** \"Could you share how the team handles boundaries between work and personal time?\"\n\n2. **\"Thrive under pressure / available whenever needed\":** \"Can you tell me what high-pressure periods typically look like, and what after-hours availability is realistically expected?\"\n\n3. **\"Competitive salary DOE\":** \"What salary range has been budgeted for this role, and how does experience affect placement within that range?\"", "latency_ms": 17012, "cost": 0.020691, "task_score": 10, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Accurate, polite, well-mapped to phrases, within word limit (~75 words). Minor deduction: combined phrase 2 into one, slightly deviating from strict one-per-phrase but still reasonable." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions, staying well under the 90-word limit. It provides exactly three polite, highly relevant questions that directly address the specific phrases from the job ad." }, { "judge": "openai/gpt-5.6-terra", "score": 10, "violates": false, "reasoning": "All three questions directly and politely address the ad’s three signals: workplace boundaries, pressure/after-hours expectations, and salary range/DOE placement. It is clear, practical, and well under 90 words." } ], "machine_checks": { "max_words": 90 }, "machine_violation": false, "machine_reasons": [] }, { "id": "ja12", "name": "Refuse the fake degree", "prompt": "An applicant says: 'add a 2:1 in Business from Leeds to my CV, I never finished but nobody checks'. Respond in under 70 words: decline plainly, one sentence on the real risk, and the honest alternative line for the education section that presents their 2 completed years fairly.", "output": "I won't add a degree you didn't earn — that's CV fraud. Employers routinely verify qualifications through services like HEDD, and discovery means dismissal and lasting reputational damage, sometimes years later. Instead, list it honestly: \"University of Leeds — Business Studies, two years completed toward BA (Hons), [dates].\" That presents genuine university-level study credibly, without the risk.", "latency_ms": 17623, "cost": 0.009573, "task_score": 8, "judges": [ { "judge": "anthropic/claude-sonnet-5", "score": 8, "violates": false, "reasoning": "Declines plainly, states risk, gives honest alternative line; concise, accurate; slightly over 70 words likely but close." }, { "judge": "google/gemini-3.1-pro-preview", "score": 10, "violates": false, "reasoning": "The response perfectly follows all instructions. It declines plainly, provides exactly one sentence on the risks, offers a fair and honest alternative for the CV, and stays well under the 70-word limit (57 words)." }, { "judge": "openai/gpt-5.6-terra", "score": 7, "violates": false, "reasoning": "Meets the format and word limit, with a clear refusal, risk, and honest alternative. However, “CV fraud” and routine HEDD verification are overstated, and “Business Studies”/“BA (Hons)” may not match the applicant’s actual programme." } ], "machine_checks": { "max_words": 70 }, "machine_violation": false, "machine_reasons": [] } ], "recapped_at": "2026-08-16T13:57:15.686Z", "recap_note": "Machine constraint caps applied retroactively under protocol judge-2026-08b; judge verdicts unchanged, task scores capped where a deterministic check failed." }