ScaffoldBench

RunOpus 5.5repeat 2

Propose-select on Where ten HTTP header fields are defined

claude-opus-5-5__propose-select__research-header-fields__r2

Score

1.000

passed
design ran as designed
Cost
$1.38
Latency
1m 34s
setup 0.0s, grading 0.1s
Turns
68
Tool calls
63
Subagents
3
Tokens in / out
548k / 19k
424,284 read from cache

Did the design run?

every check held
  • Run completed: heldharness.run-completedThe environment was built, the architecture returned and grading finished, none of them throwing.
  • No API failures: heldharness.apiNo model call failed at the API.
  • Every call ended: heldharness.calls-endedEvery call ended with a result, or was stopped by the run's own limits.
  • No harness tool denied: heldharness.harness-toolsThe sandbox guard never denied a harness tool, nor a tool the session itself offered.
  • Tool servers connected: heldharness.mcp-connectedEvery tool server the harness gave a call was connected when the session started.
  • No foreign tool servers: heldharness.mcp-hermeticNo call loaded a tool server the harness did not give it.
  • Builtin tools: heldharness.builtin-toolsA check recorded with the run.
  • Structured replies arrived: heldharness.structured-outputEvery call that asked for structured output and ran to the end got it.
  • No silent fallbacks: heldharness.not-degradedNo phase fell back on a default (an empty plan, a default winner) instead of its output.
  • All three proposers ran: heldpropose-select.proposalsThree proposers ran in parallel, each from its own angle, and none was starved while another ran.
  • The selector chose: heldpropose-select.selectionThe selector picked a proposal by a valid number, read from its schema reply and not a default.
  • Each call was shown the right tools: heldpropose-select.toolsEach proposer was shown the read-only tools, the selector none, and the executor the full toolset.

Phases

Proposal 1 (policy-first): 0.0s to 45.6s, finished, 18 tool calls, $0.356
Proposal 2 (state-first): 0.0s to 1m 1s, finished, 24 tool calls, $0.475
Proposal 3 (outcome-first): 0.0s to 40.9s, finished, 21 tool calls, $0.396
Select: 1m 1s to 1m 34s, finished, 0 tool calls, $0.153

Drawn from the phase records: timing and cost per call, no tool calls or messages.

PhaseDurationCostTurnsToolsModel
Proposal 1 (policy-first)45.6s$0.3561918claude-opus-5-5
Proposal 2 (state-first)1m 1s$0.4752524claude-opus-5-5
Proposal 3 (outcome-first)40.9s$0.3962221claude-opus-5-5
Select33.0s$0.15320claude-opus-5-5

The call-by-call replay is not published for this run. Traces are large, so the repository keeps 9 of this run set’s 900 and leaves the rest out; the score, phases and final answer on this page come from the committed run record.

Runs you can replay

Record this cell yourself: npm run bench -- run --strategies propose-select --tasks research-header-fields --models claude-opus-5-5 --repeats 1

Score breakdown

  • passheader field Age
    1.00weight 0.10

    quote verified in rfc9111

  • passheader field Cache-Control
    1.00weight 0.10

    quote verified in rfc9111

  • passheader field Expires
    1.00weight 0.10

    quote verified in rfc9111

  • passheader field Last-Modified
    1.00weight 0.10

    quote verified in rfc9110

  • passheader field ETag
    1.00weight 0.10

    quote verified in rfc9110

  • passheader field Location
    1.00weight 0.10

    quote verified in rfc9110

  • passheader field Retry-After
    1.00weight 0.10

    quote verified in rfc9110

  • passheader field Vary
    1.00weight 0.10

    quote verified in rfc9110

  • passheader field Strict-Transport-Security
    1.00weight 0.10

    quote verified in rfc6797

  • passheader field Link
    1.00weight 0.10

    quote verified in rfc8288

Penalties

No penalties were applied.

Scorer metadata

corpus
{"version":2,"snapshot":"2026-10-01","hash":"86bd1bd14f66db67ec5c7686f5e47663ee99f16c49edb5b9384b8668854857dc","documents":9}
opened documents
["rfc9111","rfc9110","rfc6797","rfc8288"]
opened by role
{"primary":["rfc9111","rfc9110","rfc6797","rfc8288"]}
displayed share
{"rfc9111":0.2935,"rfc9110":0.0657,"rfc6797":0.1469,"rfc8288":0.3466,"rfc7231":0.0026,"rfc7232":0.0041}
cited count
4
verified count
4
hallucinated count
0
citation integrity
1
hallucinated urls
[]
answer block
true
answer block repaired
false
items answered
10
item count
10
field accuracy
1
grounded items
10
field misses
{}

Final answer

I picked proposal 2. All three proposals give the same RFC and section for every field, so the quotes decided it.

Results

Why the older RFCs don't count

One check is still missing: no proposal could confirm from a status page that RFC 9110, 9111, 6797 or 8288 hasn't itself been replaced since. The RFC Editor status pages returned 404, and no opened page or search result pointed to a replacement.

About the quotes: each quote is the opening sentence of its section. The pages wrap lines and put two spaces after full stops, and these quotes are joined onto one line with single spaces. If the checker compares spacing exactly, that's the one thing that could cause a mismatch.

{"answers":[
{"field":"Age","rfc":"9111","section":"5.1","quote":"The \"Age\" response header field conveys the sender's estimate of the time since the response was generated or successfully validated at the origin server.","url":"https://www.rfc-editor.org/rfc/rfc9111.html"},
{"field":"Cache-Control","rfc":"9111","section":"5.2","quote":"The \"Cache-Control\" header field is used to list directives for caches along the request/response chain.","url":"https://www.rfc-editor.org/rfc/rfc9111.html"},
{"field":"Expires","rfc":"9111","section":"5.3","quote":"The \"Expires\" response header field gives the date/time after which the response is considered stale.","url":"https://www.rfc-editor.org/rfc/rfc9111.html"},
{"field":"Last-Modified","rfc":"9110","section":"8.8.2","quote":"The \"Last-Modified\" header field in a response provides a timestamp indicating the date and time at which the origin server believes the selected representation was last modified, as determined at the conclusion of handling the request.","url":"https://www.rfc-editor.org/rfc/rfc9110.html"},
{"field":"ETag","rfc":"9110","section":"8.8.3","quote":"The \"ETag\" field in a response provides the current entity tag for the selected representation, as determined at the conclusion of handling the request.","url":"https://www.rfc-editor.org/rfc/rfc9110.html"},
{"field":"Location","rfc":"9110","section":"10.2.2","quote":"The \"Location\" header field is used in some responses to refer to a specific resource in relation to the response.","url":"https://www.rfc-editor.org/rfc/rfc9110.html"},
{"field":"Retry-After","rfc":"9110","section":"10.2.3","quote":"Servers send the \"Retry-After\" header field to indicate how long the user agent ought to wait before making a follow-up request.","url":"https://www.rfc-editor.org/rfc/rfc9110.html"},
{"field":"Vary","rfc":"9110","section":"12.5.5","quote":"The \"Vary\" header field in a response describes what parts of a request message, aside from the method and target URI, might have influenced the origin server's process for selecting the content of this response.","url":"https://www.rfc-editor.org/rfc/rfc9110.html"},
{"field":"Strict-Transport-Security","rfc":"6797","section":"6.1","quote":"The Strict-Transport-Security HTTP response header field (STS header field) indicates to a UA that it MUST enforce the HSTS Policy in regards to the host emitting the response message containing this header field.","url":"https://www.rfc-editor.org/rfc/rfc6797.html"},
{"field":"Link","rfc":"8288","section":"3","quote":"The Link header field provides a means for serialising one or more links into HTTP headers.","url":"https://www.rfc-editor.org/rfc/rfc8288.html"}
]}

Provenance of the numbers on this page

run set
scaffold-or-scale-v1
models
Haiku 4.5, Sonnet 5.5, Opus 5.5
effort
high
repeats
2
fixtures
replay
runs
900 / 900
strategy cost
$354.44
judge cost
$9.19
wall clock
9h 48m
git
fce5807b2d

Cost excludes judge calls, which are tracked separately so a scoring model never flatters or penalises an architecture. Every figure on this site is read from results/ at build time.